CN-121979837-A - Electronic equipment structure data generation method capable of being used for rapid temperature field prediction

CN121979837ACN 121979837 ACN121979837 ACN 121979837ACN-121979837-A

Abstract

The invention discloses an electronic equipment structure data generation method capable of being used for rapid temperature field prediction, which comprises the following steps of firstly defining electronic equipment structure design parameters, wherein the parameters comprise PCB size parameters, chip layout parameters, radiator fin design parameters, heat conduction silicone grease materials, design parameters, fan design parameters and the like, secondly defining a value range of each parameter according to a design domain size, generating a multi-combination method structure parameter combination through random value and constraint verification, calculating a corresponding temperature field image by using a simulation tool, converting each group of structure parameters into a multi-channel structure image based on a preset mapping rule, and enabling the image to serve as input of a temperature field prediction neural network. The invention can automatically generate large-scale structural parameter combinations and image expression forms thereof, reduce the cost of manual modeling, improve the data preparation efficiency of a temperature field prediction model, and is suitable for deep learning temperature field prediction, intelligent heat dissipation design and optimization algorithms.

Inventors

LU LONGSHENG
LIANG LANZHI
HUANG YONGCONG
HUANG LI
YANG SHU
HUANG ZEQIANG

Assignees

华南理工大学
惠州市德赛西威汽车电子股份有限公司

Dates

Publication Date: 20260505
Application Date: 20251231

Claims (10)

1. The electronic equipment structure data generation method for fast temperature field prediction is characterized by comprising the following steps: Step one, defining structural parameters of electronic equipment; Step two, defining the value range of each parameter according to the structural design domain size of the electronic equipment; generating a plurality of groups of structure parameter groups based on constraint conditions, and storing each group of structure parameters in a parameter data table; converting each group of structural parameters into a structural image with one or more channels based on a mapping rule; And fifthly, storing the structural image generated in the fourth step in a corresponding mode to serve as input data of a temperature field prediction model, and performing steady-state or transient thermal calculation on structural parameters based on a thermal simulation tool to obtain a corresponding temperature field image serving as real label data for training the neural network.
2. The method of claim 1, wherein the structural parameters of the electronic device include various types of structural parameters describing structural parameters of structural layout, geometric properties, material properties, or cooling properties of the electronic device.
3. The method of claim 1, wherein the electronic device configuration parameters include PCB size parameters, heat conductive silicone grease chip layout parameters, heat sink fin design parameters, heat conductive silicone grease materials and design parameters, fan design parameters, and cooling channel parameters.
4. The method for generating electronic device structural data usable for rapid temperature field prediction as defined in claim 1, wherein the fourth step comprises the steps of: (1) Based on the parameter range defined in the second step, randomly taking structural parameter values; (2) Carrying out validity judgment on the structural parameter values obtained in the step (1) item by item according to the geometric constraint checking rule and the conflict judging rule; The geometric constraint in the geometric constraint checking rule should meet the space constraint of the design domain of the electronic equipment, wherein the geometric constraint comprises that the structures are not overlapped, the chip layout meets the space sequence constraint, and the positions of the heat dissipation structure and the fan or the flow channel are not in conflict; The conflict judging rule is used for judging whether the non-geometric parameters meet the design requirements, including whether the heat flux density is in a definable physical range, whether the thickness of the heat conducting material meets the manufacturability requirements, whether the flow rate of the fan is adaptive to the effective channel area of the heat radiating structure, and whether the temperature of the cooling medium is in an allowable range; (3) Dividing the parameters meeting the rule of the step (2) into corresponding structure parameter groups according to types, and recording the combination into a parameter data table after each group of structure parameters meet all constraint conditions for generating a temperature field image.
5. The method for generating electronic device structural data usable for rapid temperature field prediction as defined in claim 1, wherein the fourth step comprises the steps of: firstly, establishing a two-dimensional pixel coordinate system, then sequentially reading parameters from a parameter data table obtained in the third step, and mapping structural parameters of the electronic equipment to a position area under the two-dimensional pixel coordinate system; And finally, mapping the structural parameters of different categories to different channels of the image according to the image coding, and generating the structural image corresponding to the parameters in the step three.
6. The method of claim 5, wherein each channel is configured to represent at least one of a type of structure or attribute information of a device structure.
7. The method for generating electronic equipment structural data for rapid temperature field prediction according to claim 5, wherein three-channel images are adopted to represent different types of structural parameters, wherein a first channel represents a chip layout area, a second channel represents a heat dissipation structure area, and a third channel represents a material or cooling attribute area.
8. The method for generating structural data of electronic equipment capable of being used for rapid temperature field prediction according to claim 1, wherein the structural image generated in the fourth step is stored according to fixed resolution to be used as input data of a temperature field prediction model, steady-state or transient thermal calculation is performed on structural parameters based on an existing thermal simulation tool to obtain a corresponding temperature field distribution image, and finally the structural image-temperature field image is formed to be used as real tag data for training a neural network.
9. A computer device comprising a memory and a processor, the memory being electrically connected to the processor, the memory storing a computer program, wherein the computer program, when executed by the processor, causes the processor to implement the method as claimed in any one of claims 1 to 8.
10. A computer readable storage medium storing a computer program, wherein the computer program is executed by a processor, the processor implementing the method according to any one of claims 1 to 8.

Description

Electronic equipment structure data generation method capable of being used for rapid temperature field prediction Technical Field The invention relates to the field of data processing, in particular to an input data expression method for rapidly generating a neural network of a temperature field of electronic equipment. Background Thermal design of electronic devices typically requires temperature field solutions for complex structures by numerical simulation methods such as Finite Element Analysis (FEA), finite Volume Method (FVM), or Computational Fluid Dynamics (CFD). In these methods, it is necessary to construct a complete geometric model comprising PCB, chip, heat spreader, thermally conductive interface material and heat dissipating structure, divide a large number of grids and set corresponding boundary conditions and material parameters. The existing CFD-based electronic equipment temperature field calculation method （Multi-objective optimization of a non-uniform sinusoidal mini-channel heat sink by coupling genetic algorithm and CFD model）, needs to adopt a higher grid density and can converge through multiple iterations, so that the calculation time is long, the hardware resource consumption is large, and the requirement of the electronic equipment on quick iteration design is difficult to meet. This process requires a significant amount of computing resources and time, and the user needs to be trained. With the increasing application of deep learning in the field of engineering simulation, a research thought of replacing part of numerical calculation by using a neural network appears. The existing temperature field prediction method （A physics and data co-driven surrogate modeling approach for temperature field prediction on irregular geometric domain）, driven by the cooperation of the physics and the data realizes the prediction speed faster than the traditional simulation by learning the mapping relation between the electronic structure image and the temperature distribution. However, it relies on a large number of real simulation results as training data, and these simulation temperature fields still need to be obtained by adopting a high-cost numerical solution mode, resulting in high data preparation cost and low efficiency. The rapid temperature field generation technology needs to learn the mapping relation between the key structure of the electronic device and the temperature field, so as to realize accurate and rapid temperature field generation, which needs to generate a large number of design parameter combinations and needs to convert the parameters into the input of the deep neural network. Image-to-image translation neural networks typically employ convolutional neural networks, whose inputs are three-channel or single-channel images, which require a set of methods from parametric generation to parametric image representation to achieve rapid data generation. In summary, the prior art has the following disadvantages in terms of training data construction: 1. the real simulation data has high generation cost and long time consumption; 2. the structural parameters depend on manual modeling, and an automatic construction mode is lacked; 3. the existing image coding method does not support large-scale automatic generation of multiple groups of structural image data; 4. It is difficult to meet the requirements of the deep learning model for data size and parameter diversity. Therefore, development is needed to simplify the numerical simulation calculation process and reduce the time cost and energy efficiency of the simulation calculation scheme. Disclosure of Invention The invention aims to provide a method for quickly generating random data, which is used for generating structural parameter combinations of a multi-combination method based on random value and constraint verification by defining structural design parameters and parameter value ranges of electronic equipment, calculating corresponding temperature field images by using a simulation tool, converting each group of structural parameters into multi-channel structural image data based on a preset mapping rule, improving the data preparation efficiency of temperature field prediction, and being suitable for deep learning temperature field prediction, intelligent heat dissipation design and optimization algorithms. The invention is realized at least by one of the following technical schemes. An electronic device structure data generation method for rapid temperature field prediction, comprising the following steps: Step one, defining structural parameters of electronic equipment; Step two, defining the value range of each parameter according to the structural design domain size of the electronic equipment; generating a plurality of groups of structure parameter groups based on constraint conditions, and storing each group of structure parameters in a parameter data table; converting each group of structural parameters into a st