A method of grid space division for thermal analysis of chip packaging structure under advanced technology

Abstract

The present invention proposes a method for spatial division of thermal analysis grids based on chip packaging structures under advanced processes. According to the physical model of the chip packaging structure, the structural contour is extracted, and a grid generation algorithm is used to generate a shading grid, a through grid, and a tree grid. The shading grid is used to meet the basic grid density requirements and describe the contour of the packaging structure. The through grid is used to describe the contour of the complex heat source and boundary conditions inside the structure. The tree grid is used to adaptively encrypt the heat source and its surroundings. The combination of these three types of grids realizes the adaptation of the complex structure and complex heat source distribution inside the package. In the specific use scenario of "thermal analysis and chip packaging structure", the present invention has become a highly targeted adaptive grid division scheme, which supports thermal modeling of various advanced packaging structures and complex heat source distributions. Users can use this grid method without any threshold to obtain satisfactory results.

Description

Technical Field

[0001] The invention relates to semiconductor chip packaging technology and proposes a thermal analysis grid space partitioning method based on a chip packaging structure under advanced technology. Background Art

[0002] With the continuous evolution of process technology, temperature has become the main obstacle to limiting chip operating frequency and transistor density. Excessive temperature can cause device failure inside the chip, ultimately resulting in reduced performance or even loss of functionality of the chip. A suitable thermal model is needed to evaluate chip temperature and ultimately achieve fast and efficient thermal analysis.

[0003] Thermal modeling refers to the targeted modeling of a specific object based on thermal problems. After the modeling is completed, the entire input model has been abstracted into a series of parameters related to thermal problems. Thermal analysis refers to the operation of thermal modeling and solving of a specific object based on thermal problems, and finally obtains the temperature distribution result. In this context, the meshing method plays a key role in the entire thermal analysis as a key step in establishing the thermal model.

[0004] The object of thermal modeling is not only the chip but also the packaging structure around the chip. From the perspective of thermal analysis, the chip is the main heat source, and the peripheral packaging structure mainly plays the role of heat dissipation. A grid division method is needed to spatially discretize the chip and the packaging structure, that is, to cut the entire model into individual grids.

[0005] After the grid division is completed, each grid will be given to the subsequent solver as a basic calculation unit for calculation, and finally the temperature result of each grid will be obtained. This is a relatively complete thermal analysis process.

[0006] Therefore, the way the mesh is divided will closely affect the solution quality of the entire thermal analysis. A poor mesh may even fail to converge during the solution, ultimately resulting in unrealistic results.

[0007] There are many difficulties in this direction.

[0008] First, for chip packaging models, there are multiple cross-scale structures. For example, the thickness of the chip device layer (core heat source) is several microns, and the thickness of the silicon substrate layer around it is generally no more than 100 microns, but the thickness of the peripheral packaging substrate and heat dissipation structure is usually in millimeters. The difference in thickness between the layers makes it difficult to keep the grid at the same scale, and it is easy to have a large grid next to a small grid. This sudden change in grid size will lead to a decrease in the accuracy and convergence of the solution.

[0009] Secondly, there is often more than one chip in a package model. Current advanced packaging technology already supports multiple chips being packaged together in the same external model in a variety of different stacking methods. In a multi-chip package structure, small interconnect structures such as through silicon vias (TSV) and hybrid bonding run through or connect each chip like a thin wire. It is extremely difficult to generate corresponding grids for these small structures.

[0010] Finally, the rapid iteration of process technology and the emergence of multiple chip stacking have led to an increase in the number of heat sources in the chip, and the distribution of heat sources has become more complex. Heat accumulation often occurs in certain locations of the package model, and the heat conduction path becomes elusive. The grid should restore the most realistic distribution of heat as much as possible, distinguish between key areas and non-key areas, and make targeted adjustments.

[0011] Given the above difficulties, drawing meshes for chips and their packaging structures will be a huge challenge. Currently, there are few meshing methods that are completely tailored to chip packaging structures, and there are more general meshing algorithms. Some of them are listed below:

[0012] Icepak, a general-purpose thermal analysis software from ANSYS, has a built-in general meshing method suitable for various structures.

[0013] MENTOR (now acquired by Siemens) has a general thermal analysis software Flotherm, which has a built-in general meshing method suitable for various structures.

[0014] HOTSPOT, an academic thermal analysis software developed by the University of Virginia, provides a complete thermal analysis process for chips and packaging structures. They are a relatively early team to propose package-level thermal analysis, but because it is more academic in nature, it does not adapt the grid division method to the complex packaging structures in the industry and is only applicable to some basic and simple packaging models.

[0015] The grid partitioning algorithm is strongly related to the application scenario.

[0016] For the same model structure, engineers usually use different meshing schemes to achieve the best accuracy and efficiency when performing thermal analysis and electromagnetic field analysis. When analyzing the same problem, different model structure characteristics will also affect part of the meshing strategy. For example, for thermal analysis, the thermal analysis of computer chassis and the thermal analysis of chip packaging often have different characteristics in the meshing method.

[0017] Under this premise, although the general grid division strategy can generate a set of usable grids in various scenarios, if the model structure is too complex, the final result accuracy may not always be satisfactory. In most cases, users need to adjust the general grid based on their own experience to achieve satisfactory results.

[0018] The meshes of the thermal analysis software Icepak and Flotherm currently on the market are all general mesh algorithms, which cover many simulation fields and therefore cannot be optimized specifically for a certain problem or a specific type of structure. Inexperienced users cannot automatically obtain a complete set of optimized meshes.

[0019] However, due to its academic nature and early development, the academic thermal analysis software HOTSPOT is unable to make synchronous adaptation in terms of grid strategy for the current rapidly iterating advanced packaging structures and complex heat source distribution. Summary of the invention

[0020] In order to solve the above problems, the present invention proposes a method for meshing chip packaging structures under advanced processes for application scenarios of thermal analysis, which can handle meshing of complex physical structures and complex heat source distributions.

[0021] The technical solution of the present invention is as follows:

[0022] Based on the thermal analysis grid space division method of chip packaging structure under advanced technology, the structure contour is extracted according to the physical model of chip packaging structure, and three types of grids are generated by grid generation algorithm, namely, background grid, through grid and tree grid;

[0023] The shading grid is used to meet the basic grid density requirements and describe the outline of the packaging structure;

[0024] The through grid is used to describe the contours of complex heat sources and boundary conditions inside the structure;

[0025] The tree-like grid is used to adaptively encrypt the heat source and its surroundings;

[0026] The combination of these three types of grids can achieve adaptation to the complex structure and complex heat source distribution inside the package.

[0027] Specifically, the grid generation algorithm adopts a space division method, specifically, by setting internal grid lines, the structure of the grid to be divided is divided into multiple subspaces, and each subspace is a grid after the division is completed.

[0028] Specifically, the grid is a specific number of cuboid grids.

[0029] Specifically, the mesh generation algorithm includes the following steps:

[0030] S1 First, generate a shading grid: draw grid lines according to the outlines of each substructure in the package structure, and fill the large gaps in the local area with additional grid lines to generate a shading grid;

[0031] S2 Then, generate a through mesh: on the basis of the background mesh generation, continue to describe the local details and generate a through mesh;

[0032] S3 Finally, generate a tree-like grid: Generate a tree-like grid based on the first two types of grids, and encrypt the local key areas to improve the accuracy of the grid.

[0033] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0034] In the present invention, the grid is divided into three parts: the shading grid algorithm is used to meet the basic grid density requirements and describe the outline of the packaging structure, the through grid is used to describe the outline of the complex heat source and boundary conditions inside the structure, and on this basis, the tree grid is used to adaptively encrypt the heat source and its surroundings. The above three types of grid schemes make the present invention a highly targeted adaptive grid division scheme in the specific usage scenario of "thermal analysis and chip packaging structure", supporting thermal modeling for various advanced packaging structures and complex heat source distributions. Users can use this grid method without any threshold to obtain satisfactory results.

[0035] The grid method used in the present invention can meet the thermal modeling requirements while minimizing the size differences between grids. This is mainly reflected in the following points:

[0036] For the texture grid, an adaptive transition algorithm is used to add additional grid lines based on the contour lines, ensuring that the intervals between adjacent grid lines will not have a sudden change in proportion. At the same time, the grid lines are globally connected throughout the entire model, so the grids are orthogonal to each other, meeting the transition requirements.

[0037] For the through-mesh, the division of its grid lines depends on the position of the contour that needs to be described. In this part, there will indeed be a certain degree of large intervals adjacent to small intervals. However, since the local through-mesh only penetrates the inside of the background mesh, the size difference between the grids is controlled within a single background mesh, and no large grid difference will occur.

[0038] For tree-like grids, since an octal encryption algorithm is used, the grid will only be split 1-to-8 along the center line of the grid. The maximum ratio difference between each child grid and the parent grid is controlled at 2:1, so it can be well transitioned from key areas that require dense grids to non-key areas that require sparse grids. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 A schematic diagram of using the space partitioning method to generate a grid in a two-dimensional space (plane).

[0040] Figure 2 Schematic diagram of chip packaging structure.

[0041] Figure 3 The figure is a flow chart of the method of the present invention.

[0042] Figure 4 This is a diagram of the background grid structure of the present invention.

[0043] Figure 5 Schematic diagram of heat source distribution in the chip of the present invention.

[0044] Figure 6 A schematic diagram of the present invention in which through-grid lines are applied inside the grid to describe the contour of the heat source.

[0045] Figure 7 Schematic diagram of applying boundary conditions to the packaging model of the present invention.

[0046] Figure 8 This is a schematic diagram of applying through grid lines inside the grid to describe the boundary condition contour of the present invention.

[0047] Fig. 9 Encrypt the octree for the mesh.

[0048] Fig.10 This is a schematic diagram of the octal encryption application.

[0049] Fig.11 A schematic diagram of the encryption process.

[0050] Fig.12 This is a comparison chart of grid transition methods.

[0051] Fig.13 Schematic diagram of the hierarchical relationship between grids of the present invention.

[0052] Fig.14 It is the connection between the hierarchical relationship between the grids and the spatial position of the grids in the present invention.

[0053] Fig.15 It is a schematic diagram of the transition between the key area and the non-key area of ​​the grid in the embodiment.

[0054] Fig.16 The grid transition and actual simulated temperature distribution around four chips (two high-power chips on the top and two low-power chips on the bottom) inside the embodiment package.

[0055] Fig.17 Schematic diagram of temperature distribution of chip packaging model in the embodiment. DETAILED DESCRIPTION

[0056] The present invention proposes a method for partitioning the thermal analysis grid space of a chip packaging structure under advanced technology. According to the physical model of the chip packaging structure, the structure contour is extracted, and a grid generation algorithm is used to generate three types of grids, namely, a shading grid, a through grid, and a tree grid.

[0057] The shading grid is used to meet the basic grid density requirements and describe the outline of the packaging structure;

[0058] The through grid is used to describe the contours of complex heat sources and boundary conditions inside the structure;

[0059] The tree-like grid is used to adaptively encrypt the heat source and its surroundings;

[0060] The combination of these three types of grids can achieve adaptation to the complex structure and complex heat source distribution inside the package.

[0061] The technical solution provided by the present application will be further described below in conjunction with specific embodiments and accompanying drawings. The advantages and features of the present application will become more apparent with the following description.

[0062] Part I: Mesh Generation Algorithms

[0063] The grid generation algorithm of the present invention selects a space division method, and the grid is a rectangular parallelepiped grid.

[0064] The space division method refers to: by setting internal grid lines, the structure of the grid to be divided is divided into multiple subspaces, and each subspace is a grid after the division is completed. Figure 1 Demonstrates the use of space partitioning methods to generate meshes in two-dimensional space (plane).

[0065] The space division method has a natural advantage for square and regular structures. Although the structure targeted by the present invention is a chip packaging model, which contains a large number of tiny structures, the angles corresponding to all vertices in all structures are 90 degrees, so they can be divided into a specific number of rectangular grids. Under the above conditions, it is a relatively reasonable choice to use the space division method to generate grids. Figure 2 The specific chip packaging structure is shown. Figure (a) is the physical structure (front view), and the red in Figure (b) is the heat source distribution.

[0066] The grid generation process of the present invention is as follows Figure 3 As shown in the figure, the generated grids are mainly divided into three categories, namely, texture grids, through grids and tree grids. The combination of these three types of grids can realize the adaptation to the complex structure and complex heat source distribution inside the package.

[0067] S1 is first a background grid, the grid lines of which come from the contour lines of each substructure in the packaging structure and the supplementary grid lines for large gaps.

[0068] Each substructure has its own material, for example, Sink, Sink_base, Attrib_xacis_show and Pcb are the shells of the packaging structure, and their materials are generally copper; ChipSocket and Substrate are the substrate or base of the chip, and their materials are generally some metals; bumps, Underfill, Info_RDL, Beol are the contacts or interconnects inside the chip, and their materials are generally mixed materials; Tim and Molding are some filling media; Silicon's material is silicon, etc. The purpose of using the substructure outline is to ensure that each grid belongs to only one substructure, and there will not be a situation where a grid spans multiple substructures, thereby avoiding the existence of multiple physical materials inside a grid. But this is not enough. Simply using the outline of the substructure as the grid line will result in large grid gaps in local areas, and these large gaps need to be filled. Specifically, the present invention fills by supplementing the grid lines.

[0069] Figure 4 In the figure, the top part shows the outline of the substructure only (black dashed line), and the bottom part shows the combination of the substructure outline (black dashed line) and the supplementary grid lines (red dashed line).

[0070] S2 is followed by the through mesh, which is generated on the basis of the background mesh after it is generated. Therefore, the through mesh focuses more on the description of local details. The following introduces several situations where the through mesh appears:

[0071] Inside the S2.1 chip device layer, additional grids are required to describe the contours of the heat source.

[0072] There are many different functional modules inside the chip device layer. According to Joule's law, these functional modules will generate a lot of heat when working. These functional modules are the heat sources in thermal analysis and are usually added to the corresponding structure in a planar manner.

[0073] However, the shading grid only draws grid lines according to the outline of the substructure and cannot accurately describe the outline of the heat source. A through grid is needed to further describe the outline of the heat source.

[0074] Figure 5 In the figure, the six red substructures on the top are the device layers of the chip, and the heat sources are distributed inside them. The lower left shows the possible distribution of heat sources in a single device layer, and the lower right shows the mismatch between the grid lines (black dashed lines) and the heat sources (orange blocks).

[0075] The logic of generating the through grid is to add extra grid lines inside certain grids to describe the local contour. Here, certain grids refer to grids that intersect with the contour to be described.

[0076] Figure 6 In the figure, there is only a shading grid (black dashed line) in the upper right corner, which cannot fully describe the orange heat source boundary. Through grid lines (red dashed lines) are added below to fully describe the heat source boundary while controlling the number of grids. The through grid lines are applied to all the shading grids that intersect with the heat source outline. Figure 6 There are a total of 10 background grids formed by black dotted lines below. The through grid lines are only generated in the 6 middle background grids that intersect with the heat source. All red dotted lines are completely connected inside these 6 grids. Finally, the 6 middle background grids are transformed into 25 through grids (6+3+6+4+2+4) after the through algorithm.

[0077] S2.2 Boundary condition application layer, additional mesh is required to describe the contour of the boundary condition

[0078] At the boundary of the package model, various boundary conditions need to be applied to make the calculation closer to the actual situation. In actual usage scenarios, heat dissipation is often added to the top surface of the package, such as adding a fan to enhance air flow to increase convection heat dissipation, increasing the heat dissipation area of ​​the top copper tube to increase heat dissipation efficiency, etc. The bottom of the package is generally a large circuit board (PCB), whose area is much larger than the package structure. Due to the huge difference in area, the heat from the package will hardly affect the temperature of the circuit board (PCB), so the bottom surface of the package can be regarded as setting a constant temperature.

[0079] All the above-mentioned situations can be abstracted into a boundary condition. The boundary condition is applied in a similar way to a heat source and is also added to the corresponding structure in a plane.

[0080] Figure 7 In the figure, the upper Sink layer and PCB layer are both possible locations for applying boundary conditions. The boundary conditions will be attached to the upper surface of the Sink or the lower surface of the PCB in the form of a plane. The lower left shows the possible application of boundary conditions, and the lower right shows the mismatch between the grid lines (black dashed lines) and the boundary conditions (blue blocks).

[0081] Figure 8 In the figure, there is only a background grid (black dashed line) in the upper right corner, which cannot fully describe the blue boundary condition contour. The through grid line (red dashed line) is added below, which can fully describe the boundary condition contour while controlling the number of grids.

[0082] The through grid lines are applied to all the shading grids that intersect with the boundary condition contours. Figure 8 There are a total of 16 background grids formed by black dotted lines below. The through grid lines are only generated in the 4 background grids in the upper left corner that intersect with the boundary conditions. All red dotted lines are completely connected inside these 4 grids. Finally, the 4 background grids in the upper left corner are transformed into 9 through grids (1+2+2+4) after the through algorithm.

[0083] The last part of S3 is tree grid, which is generated based on the first two grids. Unlike through grid, tree grid is responsible for encrypting the local key areas to improve the accuracy of the grid.

[0084] Before generating the tree grid, it is necessary to first identify which locations are the key areas. In thermal analysis, the heat source is the source of heat generation and is also the key to the final numerical results. The present invention highlights the importance of the heat source in the model through the tree grid. The following are the algorithm details (existing methods):

[0085] The tree grid is represented by the octree-based encryption of the background grid and the through grid in a specific area.

[0086] Fig. 9 The figure shows an octant of a grid (the grid is divided into eight, hence the name octant).

[0087] In practical applications, the location of the heat source is first found, and then tree encryption, that is, octal encryption, is performed based on the location of the heat source.

[0088] Fig.10 In the figure, the right picture shows the area around the heat source where the background grid and the through grid have been generated, and the left picture shows the way to encrypt the area near the heat source. The black solid line is the generated background grid and the through grid, and the (black, red, green) dotted lines are the encrypted grid lines.

[0089] All meshes that are in contact with the heat source will be encrypted.

[0090] Fig.11 In the figure, the detailed encryption process is shown from left to right. The black dotted line is the first encryption, the red dotted line is the second encryption, and the green dotted line is the third encryption.

[0091] As the number of encryption times increases, the grid inside and around the heat source becomes denser and denser. Therefore, the number of encryption times can be set in advance to achieve adaptive encryption of the area around the heat source to improve the accuracy of the final thermal analysis.

[0092] The present invention uses shading grids to describe structural details and fill in large grid gaps; uses through grids to describe detailed boundaries of heat sources and boundary conditions; and uses tree grids to encrypt the core heat source and its surrounding areas to improve accuracy. The combination of the three types of grids constructs a set of adaptive grid generation algorithms for chip packaging structure and thermal analysis.

[0093] The adaptability here means that through the three types of grids mentioned above, it is possible to read only the input structure, heat source and boundary condition information without manual adjustment, and automatically generate a set of grids that meet the requirements of accuracy and efficiency.

[0094] Specifically, the background grid will automatically capture the contours of the input structure and fill in large gaps; the through grid will automatically capture the contours of heat sources and boundary conditions to achieve automatic detailed description of them; finally, the tree grid will identify the areas that need to be encrypted through the information of the heat source and automatically implement the tree encryption algorithm.

[0095] Part II: Mesh Transition Algorithm

[0096] Mesh transition means that the sizes of adjacent meshes should be close. When the mesh size fluctuates too much in a small range, the performance of the physical quantities in the discretized computational domain usually no longer conforms to the actual laws, which will aggravate the morbidity of the solution matrix and ultimately affect the accuracy and stability of the solution.

[0097] Fig.12 The figure shows a good mesh transition (left) and a poor mesh transition (right).

[0098] The grid method adopted by the present invention can reduce the size difference between grids as much as possible while meeting the thermal modeling requirements.

[0099] For the background grid, an adaptive transition algorithm is used to add supplementary grid lines based on the contour lines, ensuring that the intervals between adjacent grid lines will not have a sudden change in proportion. At the same time, the grid lines are globally connected throughout the entire model, so the grids must be orthogonal to each other to meet the transition requirements.

[0100] For the through-mesh, the division of its grid lines depends on the position of the contour that needs to be described. In this part, there will indeed be a certain degree of large intervals adjacent to small intervals. However, since the local through-mesh only penetrates the inside of the background mesh, the size difference between the grids is controlled within a single background mesh, and no large grid difference will occur.

[0101] For tree-like grids, an octo-encryption algorithm is used. The grid is split only along the center line of the grid in a 1:8 operation. The maximum ratio difference between each child grid and the parent grid is controlled at 2:1. Therefore, it is possible to transition well from key areas that require dense grids to non-key areas that require sparse grids.

[0102] Part III: Grid Data Structure

[0103] The background grid, through grid and tree grid are all in the shape of rectangular blocks, but there is a hierarchy or sequence relationship between them.

[0104] Specifically, it can be understood that there is a background grid first, and then a through grid based on the background grid. The tree grid cannot exist independently either. It must be split from the background grid, the through grid, or even itself.

[0105] In terms of hierarchical relationship, the texture grid is at the top layer (depth 1), its child grid is the through grid (depth 2), and the tree grid can be split from the through grid, the texture grid, or even from itself (depth greater than or equal to 2).

[0106] The relationship between grids can be stored through a multi-branch tree. A multi-branch tree refers to a data organization method in which each data (grid) is regarded as a node, and the connection relationship between nodes reflects the relationship between data (grids).

[0107] The main feature is that there is a root node in the multi-branch tree, and each node has only a limited number of child nodes or no child nodes.

[0108] Fig.13 It shows the hierarchical relationship between grids in the left case. In the figure, the root node (red) has 4 child nodes (black solid), and each child node (black solid) itself has 4 child nodes (black hollow).

[0109] The multi-branch tree structure composed of the hierarchical grid structure is naturally suitable for determining the spatial relationship between subsequent grids, because grids with similar spatial positions are always placed together, and the spatial relationship between the grid and its surrounding grids is synchronously established when the tree is generated.

[0110] Fig.14 In , the tree grid ABCD is split from the same through grid (grid 1), they are close in spatial position and organized under the same node. However, the tree grid Z and grid ABCD do not have the same through grid, they are far away in spatial position and are also placed far away in the data structure.

[0111] Application Examples

[0112] The overall mesh generation process is as follows: Figure 3 As shown, the physical model information (including packaging structure information and heat source information) is input externally.

[0113] The grid algorithm in the present invention first extracts the outline of the physical structure to generate the shading grid, then generates the through grid according to the input heat source and the outline information of the boundary conditions, and finally encrypts the tree grid near each heat source according to the distribution information of the heat source. In the process of generating these three types of grids, the corresponding grid data structure is also established.

[0114] The actual mesh generation algorithm works like this Fig.15 , 16 As shown in 17, as an embodiment, for example and not limitation, the post-processing software used is Paraview.

[0115] The above description is only a description of the preferred embodiments of the present application, and is not intended to limit the scope of the present application. Any changes or modifications made by any person skilled in the art based on the above disclosed technical contents shall be deemed as equivalent effective embodiments and shall fall within the scope of protection of the technical solution of the present application.

Claims

1. A thermal analysis grid space partitioning method for chip packaging structure under advanced technology, characterized in that: According to the physical model of the chip packaging structure, the structural contour is extracted, and the mesh generation algorithm is used to generate three types of meshes, namely, the background mesh, the through mesh and the tree mesh. The shading grid is used to meet the basic grid density requirements and describe the outline of the packaging structure; The through grid is used to describe the contours of complex heat sources and boundary conditions inside the structure; The tree-like grid is used to adaptively encrypt the heat source and its surroundings; The combination of these three types of grids enables adaptation to the complex structure and complex heat source distribution inside the package; The grid generation algorithm: uses a space division method to generate a grid, specifically, by setting internal grid lines, the structure of the grid to be divided is divided into multiple subspaces, each subspace is a grid after the division; The grid is a specific number of cuboid grids; The mesh generation algorithm includes the following steps: S1 First, generate a shading grid: draw grid lines according to the outlines of each substructure in the package structure, and fill the large gaps in the local area with additional grid lines to generate a shading grid; S2 Then, generate a through mesh: on the basis of the background mesh generation, continue to describe the local details and generate a through mesh; S3 Finally, generate a tree-like grid: Generate a tree-like grid based on the first two types of grids, and encrypt the local key areas to improve the accuracy of the grid.

2. The method according to claim 1, characterized in that In step S1, The shading grid follows the substructure outline, each substructure has its own material, and each grid belongs to only one substructure.

3. The method according to claim 1, characterized in that Step S2 includes: S2.1 Inside the chip device layer, add additional grids to describe the contours of the heat source; Inside the chip device layer, there are different functional modules, which are heat sources in thermal analysis and are added to the corresponding structure in a planar manner; additional through grid lines are added inside the shading grid that intersects with the heat source contour to be described to describe the local contour, and the heat source boundary is fully described while controlling the number of grids; S2.2 Boundary condition application layer, additional mesh is added to describe the contour of the boundary condition; Various boundary conditions are applied to the boundary positions of the package model; heat dissipation is added to the top surface of the package; and the bottom surface of the package is regarded as setting a constant temperature; the boundary conditions are added to the corresponding structure in a plane manner; through grid lines are added to the background grid that intersects with the boundary condition contour, and the boundary condition contour is fully described while controlling the number of grids.

4. The method according to claim 1, characterized in that In step S3, Use a tree grid to highlight the importance of heat sources in the model: The tree grid performs octree-based encryption on the background grid and the through grid in a specific area. First, the location of the heat source is found, and then the grids in contact with the heat source are tree-encrypted according to the location of the heat source, that is, octree encryption is performed. As the number of encryption times increases, the grid inside and around the heat source becomes denser and denser. By setting the number of encryption times in advance, adaptive encryption of the area around the heat source can be achieved to improve the accuracy of the final thermal analysis.

5. The method according to claim 1, characterized in that The background grid, the through grid and the tree grid have a hierarchical or sequential relationship with each other; specifically: First there is the shading grid, then there is the through grid based on the shading grid, and the tree grid cannot exist independently; its hierarchical relationship is: The texture grid is at the top layer, with a depth of 1; The through grid is a subgrid of the background grid, with a depth of 2; The tree grid is split from the through grid or the background grid, or split from itself, and the depth is greater than or equal to 2; The relationship between grids is stored through a multi-branch tree. Each grid is regarded as a node, and the connection relationship between nodes reflects the relationship between grids.

Images