CN-121997549-A - Heat dissipation assembly type selection method and system based on three-dimensional simulation and PINN collaborative optimization

CN121997549ACN 121997549 ACN121997549 ACN 121997549ACN-121997549-A

Abstract

The invention discloses a heat radiation component model selection method and system based on three-dimensional simulation and PINN collaborative optimization. The method solves the problems that the existing radiator type selection method needs a large amount of calculation resources and time and is low in precision. The method comprises the steps of constructing a three-dimensional model of a heat radiation system comprising a heat radiation component, simulating the three-dimensional model according to different working conditions of the heat radiation component, obtaining working conditions and simulated output data, establishing sample data, constructing a physical information neural network model, training to obtain a heat radiation performance prediction model, inputting working condition data of the heat radiation component to be tested into the heat radiation performance prediction model to obtain heat radiation performance prediction data, and selecting an optimal heat radiation component from comprehensive heat radiation performance, cost and space limiting factors. The invention combines the three-dimensional simulation with the physical information neural network training, and can accurately and efficiently select the radiator components meeting the heat dissipation requirement.

Inventors

ZHAO XIAOFEI
JIN ZHONGXUAN
WANG QIAOGANG

Assignees

杭州先导热电科技有限公司

Dates

Publication Date: 20260508
Application Date: 20251224

Claims (10)

1. The heat radiation assembly type selection method based on three-dimensional simulation and PINN collaborative optimization is characterized by comprising the following steps of: constructing a three-dimensional model of a heat radiation system comprising a heat radiation component, and simulating the three-dimensional model according to different working conditions of the heat radiation component; acquiring working conditions and simulation output data to establish sample data; Constructing a physical information neural network model, training the model based on sample data, and obtaining a heat radiation performance prediction model after training; inputting working condition data of the heat dissipation assembly to be tested into a heat dissipation performance prediction model to obtain heat dissipation performance prediction data; and selecting the optimal heat dissipation assembly by combining heat dissipation performance, cost and space limiting factors.
2. The method for selecting a heat dissipation assembly based on three-dimensional simulation and PINN co-optimization according to claim 1, wherein the constructing a three-dimensional model of a heat dissipation system including a heat dissipation assembly comprises the steps of: According to the structural size of the heat radiation system, constructing a three-dimensional model comprising a heat source, a heat radiation component and an air domain, wherein the heat radiation component comprises a fan and heat radiation fins; setting simulation parameters and boundary conditions; And establishing a three-dimensional simulation control equation.
3. The heat sink component model selection method based on three-dimensional simulation and PINN co-optimization according to claim 2, wherein the three-dimensional simulation control equation includes: a continuity equation, wherein the divergence of the partial derivative of air density with respect to time plus the product of air density and velocity vector is equal to 0; A momentum equation, the product of the air density and the first cell being equal to the sum of the negative value of the pressure gradient and the divergence of the second cell, the product of the air density and the gravitational acceleration vector; Wherein the first unit is the partial derivative of the velocity vector with respect to time plus the product of the velocity vector and the velocity vector gradient; the second unit is the sum of the velocity vector gradient and the velocity vector gradient transposition multiplied by the dynamic viscosity; the energy equation, the product of air density and specific heat capacity, the third unit, is equal to the sum of the divergence of the fourth unit and the heat source term; wherein the third unit is the partial derivative of temperature versus time plus the dot product of the velocity vector and the temperature gradient; the fourth element is the product of the thermal conductivity and the temperature gradient.
4. The method for selecting a heat dissipation assembly based on three-dimensional simulation and PINN co-optimization according to claim 3, wherein the simulating the three-dimensional model according to different working conditions of the heat dissipation assembly comprises: The method adopts a finite volume method to discrete a three-dimensional simulation control equation, a time item adopts a first-order implicit format, a convection item adopts a second-order windward format, a diffusion item adopts a central differential format, pressure-speed coupling adopts a SIMPLE algorithm, and an iteration convergence criterion is that all residuals are smaller than 1 x 10 -6 .
5. The heat dissipation assembly type selection method based on three-dimensional simulation and PINN co-optimization according to claim 2, 3 or 4, wherein the heat dissipation assembly type selection method is characterized in that: The working conditions of the heat dissipation assembly are obtained, wherein the working conditions comprise the fan rotating speed, the number of blades, the fin height, the fin spacing and the fin thickness; obtaining performance parameters corresponding to the working condition of the radiating component according to the simulation result, wherein the performance parameters comprise the highest temperature of the surface of the radiating fin, the lowest temperature of the surface of the radiating fin, the average temperature of the surface of the radiating fin, the temperature of an air outlet and the radiating efficiency; and preprocessing the working condition and the performance parameter of the heat radiation component to generate sample data.
6. The method for selecting a heat sink assembly based on three-dimensional simulation and PINN co-optimization of claim 5, wherein the physical information neural network model comprises: The input layer comprises 8 neurons which respectively correspond to the working condition and the air parameter of the input radiator component; The hidden layer is provided with 3 layers which respectively comprise 32,64,32 neurons, a ReLU activation function is adopted, and all the layers are in a full connection mode; The output layer comprises 5 neurons and outputs corresponding performance parameters respectively; And setting a model comprehensive loss function.
7. The method for selecting a heat sink assembly based on three-dimensional simulation and PINN co-optimization of claim 6, wherein the integrated loss function comprises a data loss and a physical loss, and a weighted sum of the data loss and the physical loss; The data loss adopts mean square error, and the physical loss is obtained by inputting the prediction result of the physical information neural network model into a control equation to calculate residual errors.
8. The method for selecting a heat sink assembly based on three-dimensional simulation and PINN co-optimization according to claim 7, wherein the training the model based on the evolution data to obtain the heat sink performance prediction model includes: Dividing sample data into a training set and a testing set, training and testing a physical information neural network model, adopting a comprehensive loss function, carrying out back propagation through an Adam optimizer, iteratively optimizing model parameters, minimizing the comprehensive loss function, and obtaining a heat radiation performance prediction model after training and testing.
9. The heat dissipation assembly type selection method based on three-dimensional simulation and PINN co-optimization according to claim 8, wherein: inputting different working conditions of the heat radiation assembly into a heat radiation performance prediction model to obtain heat radiation performance prediction data of the corresponding heat radiation assembly; according to the performance prediction data, a heat dissipation assembly with lower cost and smaller occupied space is selected on the premise of meeting the heat dissipation requirement.
10. A heat sink assembly model selection system based on three-dimensional simulation and PINN co-optimization, implementing the method as claimed in any one of claims 1-9, comprising: The simulation unit is used for constructing a three-dimensional model of the heat radiation system comprising the heat radiation component, and simulating the three-dimensional model according to different working conditions of the heat radiation component; The data acquisition processing unit acquires working conditions and simulation output data to establish sample data; The neural network unit is used for constructing a physical information neural network model, training the model based on sample data, and obtaining a heat radiation performance prediction model after training; and the selecting unit is used for deploying a heat radiation performance prediction model, inputting working condition data of the heat radiation component to be detected to obtain heat radiation performance prediction data, and selecting an optimal heat radiation component by integrating heat radiation performance, cost and space limiting factors.

Description

Heat dissipation assembly type selection method and system based on three-dimensional simulation and PINN collaborative optimization Technical Field The invention relates to the technical field of heat radiation equipment type selection, in particular to a heat radiation component type selection method and system based on three-dimensional simulation and PINN collaborative optimization. Background With the increasing power density of electronic devices, heat dissipation issues become a critical factor limiting their performance and reliability. The fan and the radiating fins are used as common radiating components, and the reasonable selection of the fan and the radiating fins is important for improving the radiating efficiency. The traditional model selection method mainly depends on an empirical formula and experimental tests, and has the problems of low precision, long period, high cost and the like. For example, in some high-power electronic devices, due to improper selection of the heat dissipation component, the temperature of the device is too high during long-time operation, so that performance is reduced, and service life is shortened. In recent years, with the development of computer technology and numerical simulation methods, three-dimensional simulation technology has been widely used in the field of heat dissipation. By establishing a three-dimensional model of the heat radiation system, the heat radiation performance under different working conditions can be simulated and analyzed, and a certain reference is provided for the type selection of the heat radiation assembly. However, the conventional three-dimensional simulation method often requires a lot of computing resources and time, and for a complicated heat dissipation system, the accuracy of the simulation result still needs to be improved. If the patent number is 202311691704.3, the model selecting method for the electric automobile radiator is named, the working condition is that the cooling fluid flow of the cooling system, the required air quantity and the air speed of the cooling system are calculated, the heat dissipation quantity is detected according to the cooling fluid flow and the air speed, and the model selecting is carried out after the electric heat dissipation quantity and the heat dissipation quantity are compared, but the model selecting method has the problems that the calculation of the cooling fluid flow, the required air quantity and the air speed of the cooling system is carried out on each radiator during model selecting, a large amount of calculation resources are occupied, and the cycle is long. Disclosure of Invention The invention mainly solves the problems that the existing radiator type selection method needs a large amount of calculation resources and time and is low in precision, and provides a radiator assembly type selection method and system based on three-dimensional simulation and PINN collaborative optimization. The technical problems are mainly solved by the following technical scheme that the heat dissipation assembly model selection method based on three-dimensional simulation and PINN cooperative optimization comprises the following steps: constructing a three-dimensional model of a heat radiation system comprising a heat radiation component, and simulating the three-dimensional model according to different working conditions of the heat radiation component; acquiring working conditions and simulation output data to establish sample data; Constructing a physical information neural network model, training the model based on sample data, and obtaining a heat radiation performance prediction model after training; inputting working condition data of the heat dissipation assembly to be tested into a heat dissipation performance prediction model to obtain heat dissipation performance prediction data; and selecting the optimal heat dissipation assembly by combining heat dissipation performance, cost and space limiting factors. According to the invention, three-dimensional simulation and PINN (physical information neural network) training are combined and applied, PINN is trained through simulation data to obtain a heat radiation performance prediction model, and the performance parameters of each heat radiation group are obtained based on the heat radiation performance prediction model, so that the heat radiator component meeting the heat radiation requirement can be accurately and efficiently selected. The model selection can be carried out by acquiring the performance parameters of the heat dissipation assembly only according to the trained heat dissipation performance prediction model, modeling is not needed to be carried out on each heat dissipation assembly, the workload and the working time are reduced, the cost is reduced, and the problems of low precision, long period and high cost of the existing heat dissipation assembly model selection method are solved. As a preferred solution, the construct