Cold plate design optimization method based on radial basis function and storage medium

Abstract

The invention relates to a cold plate design optimization method based on a radial basis function and a storage medium. The method comprises the following steps: S11, determining a first experimental point and a second experimental point of an input variable based on a Latin hypercube method, respectively performing simulation calculation on the first experimental point and the second experimental point to obtain a first data point and a second data point of a corresponding output variable, and taking the first experimental point and the first data point as a training set, and taking the second experimental point and the second data point as a test set. And S12, training the RBF-based proxy model according to the training set, judging whether a training prediction result meets a preset training requirement or not, and if yes, executing the step S13. And S13, testing the trained proxy model according to the test set, and judging whether a test prediction result meets a preset test requirement, and if so, executing the step S20. And S20, performing cold plate design based on the proxy model to obtain output variable data of the cold plate. According to the method, the cold plate can be quickly designed in the actual operation process, and the design result of the cold plate can be quickly responded.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a cold plate design optimization method based on radial basis function and a storage medium. Background Art

[0002] With the advancement of technology, the power density of electronic devices continues to increase, and the requirements for heat dissipation systems are also getting higher and higher. Traditional heat dissipation methods such as air cooling and liquid cooling can no longer meet the heat dissipation requirements of high heat flux density in some applications. Therefore, cold plate technology has emerged as an efficient heat transfer solution. The cold plate conducts heat from the heat source to the external heat dissipation system through the internal fluid channel, thereby achieving efficient heat dissipation.

[0003] In the related technologies, there are disadvantages such as a long optimization process. The design of cold plates, such as the design of skived cold plates, usually requires multiple attempts and simulation analysis, and the response is slow. For example, the adjustment of the flow channel width requires multiple attempts and simulation analysis, and the process is relatively cumbersome. Summary of the invention

[0004] The technical problem to be solved by the present invention is to provide a cold plate design optimization method and a storage medium based on radial basis functions.

[0005] The technical solution adopted by the present invention to solve the technical problem is: a cold plate design optimization method based on radial basis function, comprising:

[0006] S10, determining the input variables and output variables of the cold plate;

[0007] S11, determining a first experimental point and a second experimental point of the input variable based on the Latin hypercube method, and performing simulation calculations on the first experimental point and the second experimental point respectively to obtain a first data point and a second data point of the corresponding output variable, and using the first experimental point and the first data point as a training set, and using the second experimental point and the second data point as a test set;

[0008] S12, training the RBF-based proxy model according to the training set to obtain a training prediction result, and determining whether the training prediction result meets a preset training requirement, and if so, executing step S13;

[0009] S13, testing the trained proxy model according to the test set to obtain a test prediction result, and determining whether the test prediction result meets the preset test requirements, if so, executing step S20;

[0010] S20, designing a cold plate based on the proxy model to obtain output variable data of the cold plate.

[0011] Furthermore, in the cold plate design optimization method based on radial basis function, step S10 also includes: determining a preset parameter range of each of the input variables.

[0012] Further, in the cold plate design optimization method based on radial basis function, the simulation calculation of the first experimental point and the second experimental point in step S11 is respectively performed to obtain the first data point and the second data point of the corresponding output variable, including:

[0013] The first experimental point and the second experimental point are simulated and calculated based on the finite element method and the finite volume method to obtain the first data point and the second data point.

[0014] Further, in the cold plate design optimization method based on radial basis function, the step S12 determines whether the training prediction result meets the preset training requirement. If so, step S13 is executed, including:

[0015] It is determined whether the proxy model meets the preset fitness according to the training prediction result. If so, step S13 is executed.

[0016] Further, in the cold plate design optimization method based on radial basis function, in step S13, it is determined whether the proxy model meets the preset fit according to the training prediction result. If so, the following steps are performed:

[0017] The actual training set mean square error is calculated according to the training prediction result and the first data point of the training set, and it is determined whether the actual training set mean square error is less than the preset training set mean square error. If so, step S13 is executed.

[0018] Further, in the cold plate design optimization method based on radial basis function, the step S13 determines whether the test prediction result meets the preset test requirements. If so, step S20 is executed, including:

[0019] The actual test set mean square error is calculated based on the test prediction result and the second data point of the test set, and it is determined whether the actual test set mean square error is less than the preset test set mean square error. If so, step S20 is executed.

[0020] Furthermore, in the cold plate design optimization method based on radial basis function, the method further includes:

[0021] S30, determining the design accuracy requirement of the cold plate, and judging whether the design accuracy requirement meets the preset accuracy requirement;

[0022] S31. If yes, the cold plate is designed based on the CFD algorithm according to the input variable data of the same cold plate, and it is calculated and determined whether the error between the cold plate output variable data obtained based on the CFD algorithm and the cold plate output variable data obtained by the proxy model is within a preset error range. If yes, it is determined that the cold plate designed by the proxy model meets the high-precision requirements.

[0023] Furthermore, in the cold plate design optimization method based on radial basis function, the input variables include the scraped tooth thickness, scraped tooth interval, and working flow rate of the scraped tooth cold plate; the output variables include the thermal resistance and flow resistance of the scraped tooth cold plate.

[0024] Furthermore, in the cold plate design optimization method based on radial basis function, after step S11 and before step S12, it also includes:

[0025] S14, performing data normalization processing on the training set and the test set to obtain normalized training set and test set;

[0026] Step S12 includes:

[0027] The RBF-based proxy model is trained according to the normalized training set to obtain the training prediction result and perform a denormalization process on it, and it is determined whether the denormalized training prediction result meets the preset training requirements. If so, step S13 is executed;

[0028] Step S13 includes:

[0029] The trained proxy model is tested according to the normalized test set to obtain the test prediction result and perform denormalization on it, and it is determined whether the denormalized test prediction result meets the preset test requirements. If so, step S20 is executed.

[0030] In addition, the present invention also provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, wherein the computer program is suitable for being loaded by a processor to execute the steps of the cold plate design optimization method based on radial basis functions as described above.

[0031] The cold plate design optimization method and storage medium based on radial basis function of the present invention have the following beneficial effects: the cold plate can be quickly designed during the actual operation process to achieve a rapid response to the cold plate design results. The cold plate can be quickly designed through an agent model without relying on simulation software, which can effectively improve the cold plate design efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] The present invention will be further described below with reference to the accompanying drawings and embodiments, in which:

[0033] Figure 1 It is a schematic flow chart of a cold plate design optimization method based on radial basis function according to an embodiment of the present invention;

[0034] Figure 2 is a schematic flow chart of a cold plate design optimization method based on radial basis function in some embodiments of the present invention;

[0035] Figure 3 is a schematic flow chart of a cold plate design optimization method based on radial basis function in some embodiments of the present invention;

[0036] Figure 4 It is a schematic diagram of the cold plate structure provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0037] In order to have a clearer understanding of the technical features, purposes and effects of the present invention, specific embodiments of the present invention are now described in detail with reference to the accompanying drawings.

[0038] In the following description, specific details such as specific system structures, technologies, etc. are provided for the purpose of illustration rather than limitation, so as to provide a thorough understanding of the embodiments of the present invention. However, it should be clear to those skilled in the art that the present invention may be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to prevent unnecessary details from obstructing the description of the present invention.

[0039] In order to facilitate the understanding of the present invention, the present invention will be described more fully below, and the present invention can be implemented in many different forms and is not limited to the embodiments described herein. On the contrary, the purpose of providing these embodiments is to make the disclosure of the present invention more thorough and comprehensive.

[0040] It should be understood that although the terms first, second, third, etc. may be used to describe various information in the embodiments of the present invention, these information should not be limited to these terms. These terms are only used to distinguish the same type of information from each other. It is not necessarily required or implied that there is any such actual relationship or order between these entities or operations.

[0041] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art of the present invention. The terms used in the specification of the present invention herein are only for the purpose of describing specific embodiments and are not intended to limit the present invention.

[0042] Skived cold plates are advanced heat dissipation technologies that are mainly used in high-performance electronic devices, battery modules for new energy vehicles, and other high-power density devices. One of the key design aspects of skived cold plates is the design of the internal structure. The skived structure is a special flow channel design that is designed to improve heat exchange efficiency. This structure enhances the convective heat transfer effect by increasing the contact area between the fluid and the cold plate surface. Due to process limitations, skived cold plates are usually based on straight-through flow channels, and this design is sufficient to meet the heat dissipation needs in most scenarios.

[0043] Related technologies use a linear weighted summation model to calculate performance evaluation results, but this evaluation model is not flexible enough, is not accurate enough for performance evaluation, and cannot directly output thermal resistance and flow resistance parameters; moreover, it requires a large set of performance parameters, and cannot fully evaluate the impact of performance parameters on evaluation results when a small number of sample points are selected. There is a disadvantage of a long optimization process, and the adjustment of the flow channel width requires multiple attempts and simulation analysis, which is a cumbersome process; moreover, only the heat dissipation effect is used as the optimization goal, and the problem of flow resistance size is not considered. The evaluation goal is single, which may have an adverse effect on the overall cooling circuit. Therefore, it is necessary to establish an efficient skived cold plate design and optimization method that can achieve rapid design and response in actual operation, and complete the design without relying on simulation software when the accuracy requirements are low.

[0044] In a preferred embodiment, reference Figure 1 The cold plate design optimization method based on radial basis function of this embodiment includes the following steps:

[0045] S10. Determine the input variables and output variables of the cold plate.

[0046] Specifically, this step also includes determining a preset parameter range for each input variable.

[0047] That is to say, firstly, it is necessary to determine the input variables of the cold plate to be designed and the input variables and the numerical limit range of the input variables. For example, the input variables include but are not limited to the tooth thickness, tooth spacing, working flow rate, etc. of the skived cold plate, and the output variables generally include the thermal resistance, flow resistance, etc. of the skived cold plate.

[0048] S11. Determine the first experimental point and the second experimental point of the input variable based on the Latin hypercube method, and perform simulation calculations on the first experimental point and the second experimental point respectively to obtain the first data point and the second data point of the corresponding output variable, and use the first experimental point and the first data point as training sets, and use the second experimental point and the second data point as test sets.

[0049] Alternatively, in this step, simulation calculations are performed on the first experimental point and the second experimental point based on the finite element method and the finite volume method, respectively, to obtain the first data point and the second data point.

[0050] In other words, simulation calculations can be performed using software based on the finite element and finite volume methods to obtain data points of output variables corresponding to input variables.

[0051] S12, training the RBF-based proxy model according to the training set to obtain a training prediction result, and determining whether the training prediction result meets the preset training requirements, if so, executing step S13.

[0052] In some embodiments, step S12 includes: judging whether the proxy model meets the preset fitness according to the training prediction result, and if so, executing step S13. Optionally, if not, re-executing step S12, that is, re-training the proxy model.

[0053] Specifically, in this step, the actual training set mean square error is calculated based on the training prediction result and the first data point of the training set, and it is determined whether the actual training set mean square error is less than the preset training set mean square error. If so, step S13 is executed.

[0054] S13. Test the trained proxy model according to the test set to obtain the test prediction result, and determine whether the test prediction result meets the preset test requirements. If so, execute step S20.

[0055] In some embodiments, step S13 includes: calculating the actual test set mean square error based on the test prediction result and the second data point of the test set, and determining whether the actual test set mean square error is less than the preset test set mean square error. If so, executing step S20.

[0056] S20. Design a cold plate based on the agent model to obtain output variable data of the cold plate.

[0057] It can be understood that after the proxy model test is passed, the input variable parameters of the cold plate to be designed are selected in the proxy model to perform cold plate design, and the output variable data of the cold plate are obtained.

[0058] As an option, this step also includes: judging whether the obtained output variable data of the cold plate meets the preset design requirements, and if it does not meet the design requirements, reselecting the parameters.

[0059] This embodiment can perform rapid design of the cold plate during actual operation, achieve rapid response to the cold plate design result, and complete the rapid design of the cold plate through the proxy model without relying on simulation software, which can effectively improve the efficiency of cold plate design.

[0060] It can be understood that the core principle is to use the Latin hypercube method and CFD technology to build training sets and test sets, then train the RBF-based proxy model, and finally use the proxy model for rapid design. It is divided into two parts. The first part is the training and testing of the proxy model, and the second part is the rapid design and verification of the skived cold plate.

[0061] It should be noted that Latin Hypercube Design is a design method for computer experiments, which is a special form of Monte Carlo simulation. Latin Hypercube Design can capture the distribution characteristics of input variables with fewer sample points, thereby improving the efficiency and accuracy of the experiment. It can handle multiple input variables, and the distribution of each variable can be arbitrary, including asymmetric and discrete distributions. Latin Hypercube Design is relatively simple and easy to implement through computer programs.

[0062] CFD (Computational Fluid Dynamics) simulation calculation can obtain high-precision results, but it takes a long time. It is usually unacceptable in terms of time, so the concept of proxy model is introduced. The proxy model is to use a specific functional relationship to approximate a complex process, so as to quickly obtain results while ensuring accuracy. The RBF method is used to construct the proxy model in the present invention.

[0063] The RBF model, or radial basis function model, is a feedforward neural network based on function approximation theory, which can approximate any continuous function with arbitrary precision. The radial basis function approximation of the model is a neural network that uses a hidden layer of radial units and an output layer of linear units. The RBF approximation is characterized by a fairly fast training speed and a fairly compact network structure. They are very effective in approximating a wide range of nonlinear spaces.

[0064] Alternatively, surrogate models can be constructed using other methods such as response surface methodology, kriging models, and orthogonal polynomials.

[0065] In some embodiments, reference Figure 2 The cold plate design optimization method based on radial basis function of this embodiment further includes:

[0066] S30, determining the design accuracy requirements of the cold plate, and judging whether the design accuracy requirements meet the preset accuracy requirements.

[0067] S31. If so, the cold plate is designed based on the CFD algorithm according to the input variable data of the same cold plate, and it is calculated and determined whether the error between the cold plate output variable data obtained based on the CFD algorithm and the cold plate output variable data obtained by the proxy model is within a preset error range. If so, it is determined that the cold plate designed by the proxy model meets the high-precision requirements.

[0068] Optionally, after step S11 and before step S12, refer to Figure 3 , the optimization method of this implementation also includes:

[0069] S14. Perform data normalization on the training set and the test set to obtain normalized training set and test set.

[0070] Step S12 includes:

[0071] The RBF-based proxy model is trained according to the normalized training set to obtain the training prediction result and perform denormalization on it, and it is determined whether the denormalized training prediction result meets the preset training requirements. If so, step S13 is executed.

[0072] Step S13 includes:

[0073] The trained proxy model is tested according to the normalized test set to obtain the test prediction result and perform denormalization on it, and it is determined whether the denormalized test prediction result meets the preset test requirements. If so, execute step S20.

[0074] This embodiment can facilitate better learning of the proxy model by normalizing the data of the training set and the test set.

[0075] In a specific implementation, it is assumed that a skived cold plate with a certain size is often used in a certain scenario, such as Figure 4 The figure shows the appearance structure of the skived cold plate. In the figure, label 1 is the coolant inlet, label 2 is the coolant outlet, and label 3 is the skived tooth. Only three variables, namely, the skived tooth thickness, the skived tooth interval, and the working flow rate, are considered. The cold plate material is copper, and the coolant is a 25% propylene glycol-water solution.

[0076] The input variables of the cold plate and their corresponding preset parameter ranges are shown in Table 1.

[0077] Table 1

[0078] Shovel tooth thickness range / mm 0.2-1 Spacing range of shovel teeth / mm 0.2-1 Working flow range / LPM 0.5-3

[0079] The proxy model is constructed based on the data shown in Table 1. Specifically, the first experimental point and the second experimental point of the input variables of the training set and the test set are constructed using the Latin hypercube method, as shown in Table 2.

[0080] Table 2

[0081]

[0082] The thermal resistance and flow resistance corresponding to the experimental points of the above input variables can be calculated using CFD software, as shown in Table 3.

[0083] Table 3

[0084]

[0085] The RBF-based proxy model is trained using the training set, and then tested using the test set. For the case where the output variable is thermal resistance, refer to the python code as shown in the attachment. The basic steps are as follows:

[0086] 1. Import necessary libraries: Import NumPy library for mathematical operations, MinMaxScaler function for feature scaling, and mean_squared_error function for calculating mean square error (MSE).

[0087] 2. Define the training set and test set: Use NumPy arrays to define the training set and test set data.

[0088] 3. Custom RBF model: A class named CustomRBFModel is created, which contains the implementation of RBF kernel function, fitting method and prediction method.

[0089] 4. Data normalization: Use the MinMaxScaler function to normalize the features of the training set and test set to facilitate better learning of the model.

[0090] 5. Model training: Create an instance of CustomRBFModel and train it using the normalized training data.

[0091] 6. Model prediction: Use the trained model to make predictions on the training set and test set.

[0092] 7. Denormalize the prediction results: Denormalize the prediction results to restore them to the scale of the original data.

[0093] 8. Calculate MSE: Calculate and print the mean square error of the training set and test set to evaluate the fit and performance of the model.

[0094] 9. Print prediction results: Print the prediction results and actual values ​​of the test set for comparison.

[0095] The training results of the thermal resistance prediction model are shown in Table 5. Here, the mean square error is used for judgment, requiring the mean square error of the training set to be less than 5e-20 and the mean square error of the test set to be less than 0.0002. If the requirements are met, the model construction is completed. Similarly, the training of the flow resistance proxy model can be done using the same method, which will not be repeated here.

[0096] Table 4

[0097]

[0098] It can be seen that the training and testing of the proxy model meet the corresponding requirements.

[0099] Suppose you need to design a cold plate with the design requirements of 1.2 LPM flow rate, less than 0.15°C / W thermal resistance, and less than 3 kPa pressure drop. The initial design parameters are shown in Table 5.

[0100] Table 5

[0101] Shovel tooth thickness / mm 0.3 Spacing of shovel teeth / mm 0.6 Work flow / LPM 1.2

[0102] Using the proxy model for quick calculation, the thermal resistance and flow resistance can be obtained. Since higher accuracy is required, CFD calculation verification is used. The results are shown in Table 6 below.

[0103] Thermal resistance / ℃W <![CDATA[Flow resistance / kPa]]> Proxy Model 0.123 2.2 CFD 0.133 2.38 error 7.5％ 7.6％

[0104] After CFD verification, both thermal resistance and flow resistance meet the requirements and the design is completed.

[0105] As you can understand, the thermal resistance definition uses the following calculation formula to calculate the thermal resistance of the cold plate:

[0106] R=(T c -T in ) / Q

[0107] Where R is thermal resistance, unit is ℃ / W; T c is the temperature at the center of the bottom of the cold plate, in °C; T in is the coolant temperature at the cold plate inlet, in °C; Q is the heat source power, in W.

[0108] The flow resistance in this article is defined using the following calculation formula to calculate the flow resistance of the cold plate:

[0109] P=P2-P1

[0110] Among them, P is the flow resistance, the unit is kPa; P2 is the pressure at the cold plate liquid inlet joint, the unit is kPa; P1 is the pressure at the cold plate liquid outlet joint, the unit is kPa.

[0111] In this embodiment, the Latin hypercube design can evenly distribute the sample points and fully consider the influence of different performance parameters. The proxy model can directly output the thermal resistance and flow resistance values ​​to confirm whether the design requirements are met. The proxy model has a fast calculation speed and can be used for rapid design when the accuracy requirements are not high, without simulation verification.

[0112] The following example provides a proxy model training code reference:

[0113] import numpy as np from sklearn.preprocessing import MinMaxScalerfrom sklearn.metrics import mean_squared_error#Read training set data

[0114] train_data = np.array([

[0115] [0.564,0.2,1.86,0.077997143],

[0116] [0.636,0.564,2.55,0.1083],

[0117] [0.709,0.491,3,0.099477143],

[0118] [0.782,0.782,1.18,0.154191429],

[0119] [0.855,0.709,0.5,0.200742857],

[0120] [0.927,0.345,1.41,0.155334286],

[0121] [1,1,0.95,0.186485714],

[0122] [0.2,0.927,2.77,0.127265714],

[0123] [0.273,0.418,2.32,0.09566],

[0124] [0.345,0.273,0.73,0.10222],

[0125] [0.418,0.855,2.09,0.133114286],

[0126] [0.491,0.636,1.64,0.127328571],

[0127] ],dtype=float)

[0128] #Read test set data

[0129] test_data = np.array([

[0130] [0.2,0.2,3,0.0675],

[0131] [0.47,0.73,2.17,0.1062],

[0132] [0.73,0.47,0.5,0.1651],

[0133] [1,1,1.33,0.1676], ])

[0135] #Define a custom RBF model

[0136] class CustomRBFModel:

[0137] def__init__(self,gamma):

[0138] self.gamma = gamma

[0139] self.X_train=None

[0140] self.weights = None

[0141] def_rbf_kernel(self,X,Y):

[0142] # Calculate the kernel matrix between X and Y

[0143] sq_dists=np.expand_dims(np.sum(X**2,axis=1),1)+np.sum(Y**2,axis=1)-2*X@YT

[0144] return np.exp(-self.gamma*sq_dists)

[0145] def fit(self,X,y):

[0146] self.X_train = X

[0147] K = self._rbf_kernel(X,X)

[0148] # Use pinv to calculate weights

[0149] self.weights = np.linalg.pinv(K)@y

[0150] def predict(self,X):

[0151] K_star=self._rbf_kernel(X,self.X_train)

[0152] return K_star@self.weights

[0153] # Load and normalize data

[0154] scaler_x=MinMaxScaler(feature_range=(0,1))

[0155] scaler_y=MinMaxScaler(feature_range=(0,1))

[0156] #Separate and normalize the training set data

[0157] X_train,y_train=train_data[:,:-1],train_data[:,-1].reshape(-1,1)

[0158] X_train_scaled=scaler_x.fit_transform(X_train)

[0159] y_train_scaled=scaler_y.fit_transform(y_train)

[0160] #Separate and normalize the test set data

[0161] X_test,y_test=test_data[:,:-1],test_data[:,-1].reshape(-1,1)

[0162] X_test_scaled=scaler_x.transform(X_test)

[0163] y_test_scaled=scaler_y.transform(y_test)

[0164] #Create a model instance

[0165] gamma_value=2.656

[0166] model=CustomRBFModel(gamma=gamma_value)

[0167] #Train model

[0168] model.fit(X_train_scaled,y_train_scaled)

[0169] #Predict training set and test set

[0170] y_train_pred=model.predict(X_train_scaled)

[0171] y_test_pred=model.predict(X_test_scaled)

[0172] #Denormalize prediction results

[0173] y_train_pred_inv=scaler_y.inverse_transform(y_train_pred.reshape(-1,1))y_test_pred_inv=scaler_y.inverse_transform(y_test_pred.reshape(-1,1))

[0174] # Calculate and print MSE

[0175] train_mse=mean_squared_error(y_train.ravel(),y_train_pred_inv.ravel())

[0176] test_mse=mean_squared_error(y_test.ravel(),y_test_pred_inv.ravel())

[0177] print(f"Training set MSE:{train_mse}")

[0178] print(f"Test set MSE:{test_mse}")

[0179] #Print the predicted results and actual values ​​of the test set

[0180] print("Test set prediction results:",y_test_pred_inv)

[0181] print("Actual value of the test set:",y_test.ravel())

[0182] In another preferred embodiment, the computer-readable storage medium of this embodiment stores a computer program, and the computer program is suitable for being loaded by a processor to execute the steps of the cold plate design optimization method based on radial basis function as described in the above embodiment.

[0183] This embodiment can perform rapid design of the cold plate during actual operation, achieve rapid response to the cold plate design result, and complete the rapid design of the cold plate through the proxy model without relying on simulation software, which can effectively improve the efficiency of cold plate design.

[0184] The computer-readable storage medium of the present invention can be any computer-readable storage medium that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a magnetic disk, or an optical disk.

[0185] The processor of the present invention is used to provide computing and control capabilities to support the operation of the entire device. It should be understood that in the embodiments of the present application, the processor may be a central processing unit (CPU), and the processor may also be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. Among them, the general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0186] Professionals may further appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in the above description according to function. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians may use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0187] The steps of the method or algorithm described in conjunction with the embodiments disclosed herein may be implemented directly using hardware, a software module executed by a processor, or a combination of the two. The software module may be placed in a random access memory (RAM), a memory, a read-only memory (ROM), an electrically programmable ROM, an electrically erasable programmable ROM, a register, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.

[0188] It can be understood that the above embodiments only express the preferred implementation modes of the present invention, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the patent scope of the present invention. It should be pointed out that, for ordinary technicians in this field, the above technical features can be freely combined without departing from the concept of the present invention, and several deformations and improvements can be made, which all belong to the protection scope of the present invention. Therefore, all equivalent changes and modifications made to the scope of the claims of the present invention should belong to the coverage of the claims of the present invention.

Claims

1. A cold plate design optimization method based on radial basis function, characterized in that: include: S10, determining the input variables and output variables of the cold plate; S11, determining a first experimental point and a second experimental point of the input variable based on the Latin hypercube method, and performing simulation calculations on the first experimental point and the second experimental point respectively to obtain a first data point and a second data point of the corresponding output variable, and using the first experimental point and the first data point as a training set, and using the second experimental point and the second data point as a test set; S12, training the RBF-based proxy model according to the training set to obtain a training prediction result, and determining whether the training prediction result meets a preset training requirement, and if so, executing step S13; S13, testing the trained proxy model according to the test set to obtain a test prediction result, and determining whether the test prediction result meets the preset test requirements, if so, executing step S20; S20, designing a cold plate based on the proxy model to obtain output variable data of the cold plate.

2. The cold plate design optimization method based on radial basis function according to claim 1 is characterized in that: Step S10 also includes: determining a preset parameter range of each of the input variables.

3. The cold plate design optimization method based on radial basis function according to claim 1 is characterized in that: The step S11 of performing simulation calculations on the first experimental point and the second experimental point respectively to obtain the corresponding first data point and second data point of the output variable includes: The first experimental point and the second experimental point are simulated and calculated based on the finite element method and the finite volume method to obtain the first data point and the second data point.

4. The cold plate design optimization method based on radial basis function according to claim 1, characterized in that: The step S12 determines whether the training prediction result meets the preset training requirement. If so, step S13 is executed, including: It is determined whether the proxy model meets the preset fitness according to the training prediction result. If so, step S13 is executed.

5. The cold plate design optimization method based on radial basis function according to claim 4 is characterized in that: In step S13, it is determined whether the proxy model meets the preset fitness according to the training prediction result. If so, step S13 is executed, including: The actual training set mean square error is calculated according to the training prediction result and the first data point of the training set, and it is determined whether the actual training set mean square error is less than the preset training set mean square error. If so, step S13 is executed.

6. The cold plate design optimization method based on radial basis function according to claim 1, characterized in that: The step S13 determines whether the test prediction result meets the preset test requirements. If yes, step S20 is executed, including: The actual test set mean square error is calculated based on the test prediction result and the second data point of the test set, and it is determined whether the actual test set mean square error is less than the preset test set mean square error. If so, step S20 is executed.

7. The cold plate design optimization method based on radial basis function according to claim 1, characterized in that: The method further comprises: S30, determining the design accuracy requirement of the cold plate, and judging whether the design accuracy requirement meets the preset accuracy requirement; S31. If yes, the cold plate is designed based on the CFD algorithm according to the input variable data of the same cold plate, and it is calculated and determined whether the error between the cold plate output variable data obtained based on the CFD algorithm and the cold plate output variable data obtained by the proxy model is within a preset error range. If yes, it is determined that the cold plate designed by the proxy model meets the high-precision requirements.

8. The cold plate design optimization method based on radial basis function according to claim 1, characterized in that: The input variables include the skived tooth thickness, the skived tooth interval, and the working flow rate of the skived tooth cold plate; the output variables include the thermal resistance and the flow resistance of the skived tooth cold plate.

9. The cold plate design optimization method based on radial basis function according to claim 1, characterized in that: After step S11 and before step S12, the method further includes: S14, performing data normalization processing on the training set and the test set to obtain normalized training set and test set; Step S12 includes: The RBF-based proxy model is trained according to the normalized training set to obtain the training prediction result and perform a denormalization process on it, and it is determined whether the denormalized training prediction result meets the preset training requirements. If so, step S13 is executed; Step S13 includes: The trained proxy model is tested according to the normalized test set to obtain the test prediction result and perform denormalization on it, and it is determined whether the denormalized test prediction result meets the preset test requirements. If so, step S20 is executed.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and the computer program is suitable for being loaded by a processor to execute the steps of the cold plate design optimization method based on radial basis function according to any one of claims 1 to 9.

Images