Electromagnetic field simulation adaptive grid generation method taking electromagnetic field distribution characteristics as core guidance and system thereof

Abstract

The invention discloses an electromagnetic field simulation adaptive grid generation method taking electromagnetic field distribution characteristics as core guidance and a system thereof. Combining the geometric model data of each radio frequency device with the connection relation information to form a geometric representation containing topological connection information; setting a current sampling point on a non-contact surface of the surface of the metal radio frequency device and other radio frequency devices based on the geometric representation containing the topological connection information, and obtaining a current value of the current sampling point; s parameters of two rounds and current values of current sampling points are obtained through the first two rounds of mesh generation; if the convergence condition is met, the grid generated in the first round is used for full-wave-band simulation; and if the convergence condition is not met, carrying out a local encryption process until the convergence condition is met, and carrying out full-wave band simulation by using the grid generated in the previous round. According to the method, the limitation that traditional grid generation is only adaptive to geometric shapes and neglects electromagnetic characteristics is broken through, and collaborative optimization of grid quality and calculation efficiency in a high-frequency electromagnetic field simulation scene is realized.

Description

Technical Field

[0001] This invention relates to the field of electromagnetic field simulation technology, specifically to an adaptive mesh generation method and system for electromagnetic field simulation based on electromagnetic field distribution characteristics. Background Technology

[0002] In modern engineering and scientific research, electromagnetic field simulation technology is a key tool for analyzing and designing electromagnetic equipment, communication systems, microwave devices, and more. Mesh generation, as a crucial preliminary step in electromagnetic field simulation, directly determines the accuracy of the simulation results and the computational efficiency.

[0003] Existing mesh generation methods (such as structured and unstructured mesh generation) have significant limitations: these methods focus on adapting to geometry and do not fully consider electromagnetic properties. In high-frequency electromagnetic field simulations, the skin effect causes current and electromagnetic fields to concentrate in a thin layer on the conductor surface, where the electromagnetic field changes drastically and has extremely high gradients. However, existing adaptive meshing techniques adjust the mesh based on fixed error estimation criteria, making it difficult to accurately adapt to the special distribution caused by the skin effect. Either they fail to capture the rapid changes in the electromagnetic field on the conductor surface, leading to simulation results that deviate from reality; or they excessively refine the mesh in non-critical areas, consuming a large amount of computational resources and reducing simulation efficiency.

[0004] Therefore, there is an urgent need for a method that can fully consider the actual distribution characteristics of electromagnetic fields and adaptively generate high-quality meshes to improve the accuracy and efficiency of electromagnetic field simulation. Summary of the Invention

[0005] The purpose of this invention is to address the shortcomings of existing technologies by providing an adaptive mesh generation method and system for electromagnetic field simulation based on electromagnetic field distribution characteristics.

[0006] In a first aspect, the present invention provides an adaptive mesh generation method for electromagnetic field simulation based on electromagnetic field distribution characteristics, the method comprising: Obtain the geometric model data and electromagnetic characteristic parameters of each radio frequency device model in the integrated circuit; The connection relationship information of each RF device is extracted from the geometric model data of each RF device model; the geometric model data and the connection relationship information are merged to form a geometric representation containing topological connection information. The initial mesh size that can initially cover the key area is calculated based on electromagnetic property parameters; Based on the geometric representation containing topological connection information, current sampling points are set at intervals of a1 times the initial grid size on the surface of the metal RF device and the non-contact surface with other RF devices, where a1 > 0; during the simulation, the current value of the current sampling point is obtained by interpolating the vertices of the grid cells where the current sampling point is located. The S-parameters and current values ​​at the current sampling points are obtained through the first two rounds of mesh generation. If the convergence condition is met, the mesh generated in the first round is used for full-band simulation. If the convergence condition is not met, a local refinement process is performed until the convergence condition is met, and then the mesh generated in the previous round is used for full-band simulation.

[0007] Optionally, the geometric model data includes the contour point coordinates and dimensions of the radio frequency device; the electromagnetic characteristic parameters include the maximum relative permittivity of the metal layer. Simulated maximum frequency .

[0008] Optionally, the process of extracting the connection relationship information of each radio frequency device based on the geometric model data of each radio frequency device model is as follows: Generate a 3D axis-aligned bounding box for the geometric model data of each RF device; A three-dimensional KD tree spatial index structure is constructed based on the three-dimensional axis-aligned bounding boxes of all RF devices. Traverse each radio frequency device and use the three-dimensional KD tree spatial index structure to search for a list of potential contact objects that spatially overlap with their bounding boxes; Perform a geometric Boolean intersection operation on the radio frequency devices in the searched list of potential contact objects to obtain the contact surface; The obtained contact surface geometric model data is merged with the initial geometric model data of the RF device to construct a unified data structure that distinguishes between contact surfaces and non-contact surfaces, forming a geometric representation containing complete topological connection information.

[0009] Optionally, the process of obtaining S-parameters and current values ​​at current sampling points through the first two rounds of mesh generation; if the convergence condition is met, using the mesh generated in the first round for full-band simulation; if the convergence condition is not met, performing local refinement is implemented as follows: In the first round, a mesh generation algorithm is used to generate a mesh with a size a2 times the initial mesh size, where a2 > a1. The electromagnetic integral equation at the maximum frequency point of the simulation is solved, and the S-parameters and current values ​​at the current sampling points in the first round are extracted. In the second round, a mesh generation algorithm is used to generate a mesh with a size a3 times the initial mesh size, where a2 > a3 > a1. The electromagnetic integral equation at the maximum frequency point of the simulation is solved in the same way as in the first round, and the S-parameters and current values ​​at the current sampling points in the second round are extracted. If the difference ΔS between the two rounds of S-parameters is less than the threshold If convergence is achieved, the first round of meshing is used for full-band simulation; otherwise, the local refinement process begins.

[0010] Optionally, the implementation process of the local encryption procedure is as follows: S5-1: Local refinement of coordinate points in the geometric model contour, specifically: In the geometric model contour, coordinate points are divided into current sampling points and non-current sampling points; The grid size is updated by classifying current sampling points into encrypted and unencrypted points; A non-uniform subdivision size field is used to generate the geometric model contour of non-current sampling points; S5-2: Refined Mesh Generation and Convergence Judgment: Based on Non-Uniform Mesh Size Field A mesh is constructed using a mesh generation algorithm, and the S-parameters of the nth round are solved and extracted; if the difference ΔS between the S-parameters of the nth round and the (n-1th round)th round is less than a threshold... If convergence is achieved, the mesh generated in the (n-1)th round is used for full-band simulation; otherwise, return to step S5-1 to repeat local refinement.

[0011] Optionally, the process of classifying and updating the grid size by using encrypted and unencrypted points for current sampling points is as follows: Calculate the relative rate of change of current at the current sampling point The current sampling points with the largest relative rate of change of the top N% of currents are selected as the densification points; 0 < N% < 100%; if the i-th current sampling point in the nth round is a densification point, its grid size is thinned by the densification coefficient; if the i-th current sampling point in the nth round is a non-densification point, the grid size remains unchanged, and the following updates are performed: Equation (1) Where n≥3, For encryption coefficients, <1.

[0012] Optionally, the process of generating the non-uniform subdivision size field of the geometric model contour for non-current sampling points is as follows: Using the grid size of the current sampling points as the control source, a non-uniform mesh size field for generating the geometric model profile is generated using an inverse distance weighting method with attenuation. .

[0013] Secondly, the present invention provides an adaptive mesh generation system for electromagnetic field simulation, comprising: The data acquisition module is responsible for acquiring the geometric model data and electromagnetic characteristic parameters of each radio frequency device model in the integrated circuit; The first calculation module is responsible for extracting the connection relationship information of each radio frequency device based on the geometric model data of each radio frequency device model; merging the geometric model data and the connection relationship information to form a geometric representation containing topological connection information; The second calculation module is responsible for calculating the initial grid size that can initially cover the key area based on electromagnetic property parameters; The third calculation module is responsible for setting current sampling points on the surface of the metal radio frequency device at intervals of a1 times the initial grid size on the non-contact surfaces with other radio frequency devices, based on the geometric representation containing topological connection information; during the simulation, the current value of the current sampling point is obtained by interpolating the vertices of the grid cells where the current sampling point is located. The fourth calculation module is responsible for obtaining the S-parameters and current values ​​of the current sampling points from the first two rounds of mesh generation. If the convergence condition is met, the mesh generated in the first round is used for full-band simulation. If the convergence condition is not met, a local refinement process is performed until the convergence condition is met, and then the mesh generated in the previous round is used for full-band simulation.

[0014] Thirdly, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed in a computer, causes the computer to perform the method described thereon.

[0015] Fourthly, the present invention provides a computing device, including a memory and a processor, wherein the memory stores executable code, and the processor executes the executable code to implement the method described above.

[0016] The beneficial effects of this invention are at least as follows: This invention proposes an adaptive mesh generation method guided by the characteristics of electromagnetic field distribution. Through the whole process design of "electromagnetic parameter-driven initial mesh calculation - multi-round subdivision and S-parameter convergence judgment - current change rate-guided local refinement - construction of size field by inverse distance weighting method with attenuation", it breaks through the limitation of traditional mesh generation that only adapts to geometry and ignores electromagnetic characteristics, and realizes the synergistic optimization of mesh quality and computational efficiency in high-frequency electromagnetic field simulation scenarios.

[0017] The improved performance brought about by this method is mainly reflected in the following aspects: Significantly improved simulation accuracy: To address the problem of drastic changes in the electromagnetic field on the conductor surface caused by the skin effect in high-frequency scenarios, this method captures the differences in current distribution through current sampling points and filters the densified regions by the relative rate of change of current. This ensures that key areas with high electromagnetic field gradients (such as non-contact surfaces of conductors) obtain accurate local mesh densification, avoiding deviations in current and S-parameter calculations caused by overly coarse meshes. This makes the simulation results more consistent with the actual electromagnetic distribution, especially in scenarios sensitive to electromagnetic characteristics such as RF devices. The S-parameter calculation error can be controlled within the threshold, significantly reducing the deviation between simulation results and actual performance.

[0018] High computational resource utilization: Unlike traditional methods that waste resources due to "global uniform densification" or "fixed criterion densification," this method avoids over-densification in non-critical areas through multiple rounds of convergence judgment (using the S-parameter difference as the core indicator). Mesh refinement is only performed in areas with large current variations, while undensed areas maintain their original mesh size. Simultaneously, a weighted inverse distance method with attenuation is used to achieve a smooth transition in size between densified and undensed areas, reducing the number of redundant mesh cells. In practical applications, in simulation scenarios of typical high-frequency RF devices (such as microwave filters and RF antennas), compared to global densification schemes, it can reduce computational load by 20%-30%, improving simulation speed while maintaining accuracy. Attached Figure Description

[0019] To more clearly illustrate the technical solution of the present invention, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is a flowchart of an adaptive mesh generation method for electromagnetic field simulation based on electromagnetic field distribution characteristics, provided by an embodiment of the present invention.

[0021] Figure 2 This is a mesh effect diagram generated before using the method described in this invention.

[0022] Figure 3 The mesh effect image is generated after using the method described in this invention. Detailed Implementation

[0023] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0024] This implementation provides an adaptive mesh generation method for electromagnetic field simulation based on electromagnetic field distribution characteristics. See Appendix. Figure 1 This includes the following steps: Step S1: Data Acquisition and Preprocessing Obtain the geometric model data and electromagnetic characteristic parameters of each radio frequency (RF) device model in the integrated circuit; wherein, the geometric model data includes the contour point coordinates and dimensions of the RF device; and the electromagnetic characteristic parameters include the maximum relative permittivity of the metal layer. Simulated maximum frequency ; After discretizing the geometric model data of the radio frequency devices, a global collision detection algorithm based on three-dimensional spatial indexing is used to calculate the connection relationship of each radio frequency device model. The specific steps are as follows: Generate 3D axis-aligned bounding boxes: Traverse the geometric model data of all RF devices, calculate the minimum and maximum coordinates of each device in the X, Y, and Z axes, and generate the corresponding 3D axis-aligned bounding boxes.

[0025] Constructing a 3D KD tree: Using the bounding boxes aligned with all the above 3D axes as input objects, construct a 3D KD tree. The KD tree divides the space along the X, Y, and Z axes in turn, organizing the 3D space into a hierarchical structure, thereby storing spatially adjacent objects in nearby nodes.

[0026] Coarse screening of potential contact objects: For the target RF device, a range query is performed in the KD tree to retrieve all other RF devices whose bounding boxes spatially intersect with the target RF device, forming a list of potential contact candidates. This step quickly eliminates objects that are spatially far away and impossible to contact.

[0027] Contact surface solution: Perform precise geometric Boolean intersection operation only on devices in the candidate list. If an intersection exists, the two RF devices are determined to be in contact, and the contact surface is saved.

[0028] Generate topology representation: Remove all extracted contact surfaces from the initial geometric surface of the RF device, and mark the remaining area as a free surface. Finally, merge the geometric model data with the initial geometric model data of the RF device to form a geometric representation containing complete topology connection information, which is used for current sampling point setting in the subsequent step S3.

[0029] Step S2: Calculation of initial mesh size The initial mesh size (meshsize) that can initially cover the key area is calculated based on electromagnetic property parameters, using the following formula: Equation (1) Where c is the speed of light in a vacuum (3 × 10⁻⁶). 8 m / s), k is an empirical coefficient (taken as 10 in this embodiment, but can be adjusted according to the application scenario). To simulate the maximum frequency, is the maximum relative permittivity of the metal layer.

[0030] Formula (1) adapts the electromagnetic field wavelength characteristics in high-frequency scenarios by simulating the maximum frequency and the maximum relative permittivity of the metal layer, ensuring that the initial mesh can initially cover the key areas.

[0031] Step S3: Current sampling point generation and current calculation Based on the geometric representation containing topological connectivity information obtained in step S1, current sampling points are set at intervals of a1 times the initial grid size in areas on the surface of the radio frequency device made of metallic material that are determined not to be in contact with other radio frequency devices, in order to capture changes in the current distribution on the surface of the radio frequency device. In this embodiment, a1 is set to 1 (the larger the grid size, the sparser the grid).

[0032] During the simulation, the current value of the current sampling point is obtained by interpolating the vertices of the grid cells containing the current sampling point. In this embodiment, a triangular mesh is used. The interpolation process for the vertices of the mesh is as follows: Let the three vertices of the triangular mesh element be r1, r2, and r3, and the current sampling point be r0. Using the basis function expansion based on area coordinates, the current value at the current sampling point can be obtained. for: Equation (2) Equation (3) in Let r0 be the weight of the barycenter coordinates within the triangular mesh cell. The values ​​at the vertices of the triangular mesh elements are obtained by solving the electromagnetic integral equation using the method of moments in subsequent step S4, i = 1, 2, 3.

[0033] Step S4: Multi-round adaptive mesh generation and convergence judgment In multiple rounds of meshing, the position of the current sampling point remains consistent with the setting in step S3, and only the corresponding grid size and current value are updated.

[0034] Round 1 mesh generation: Delaunay triangulation is used to generate a mesh with a size a2 times the initial mesh size, where a2 > a1. The electromagnetic integral equation at the maximum frequency point of the simulation is solved based on the method of moments. The total port current at each port is extracted from the solution vector of the electromagnetic equation. Combined with the port excitation conditions, a multi-port network Y-parameter matrix is ​​constructed. Based on the preset reference impedance, the impedance matrix is ​​converted into an S-parameter matrix using the network parameter transformation formula, thus completing the first round of S-parameter extraction and current value extraction at the current sampling points. In this embodiment, a2 is set to 2.

[0035] Second round of mesh generation: Delaunay triangulation is used to generate a mesh with a size a3 times the initial mesh size, where a2 > a3 > a1. The electromagnetic integral equation at the maximum simulation frequency point is solved using the method of moments, the same as in the first round, to extract the S-parameters and current sampling points for the second round. In this embodiment, a3 is set to 1.5.

[0036] Convergence criterion: If the difference ΔS between the two rounds of S-parameters is less than the threshold τS, convergence is determined. At this point, the first round of mesh meets the accuracy requirements, and the first round of mesh is used for full-band simulation (to avoid redundant calculations of finer meshes); otherwise, proceed to step S5, the local refinement process. In this embodiment, τS is set to 0.02.

[0037] Step S5. Local encryption process (rounds n≥3) 5-1 The coordinate points in the geometric model outline are refined, specifically: In the geometric model contour, coordinate points are divided into current sampling points and non-current sampling points; The current sampling points are classified into encrypted and unencrypted points for grid size update, specifically: Calculate the relative rate of change of current at the current sampling point The current sampling points with the largest relative rate of change of current in the top N% are selected as the encryption points. In this embodiment, N% = 20%.

[0038] Equation (4) Where I2 is the current value at a certain current sampling point in the (n-1)th round, I1 is the current value at the same current sampling point as I2 in the (n-2)th round, and n≥3.

[0039] For the grid size at the current sampling point Round n Grid size The update rules are as follows: Equation (5) Where n≥3, For encryption coefficients, <1, in this embodiment the value is 0.8.

[0040] The non-uniform subdivision size field for generating the geometric model contour of non-current sampling points is specifically: Using the grid size of the current sampling points as the control source, an inverse distance weighting method with attenuation is used to generate a non-uniform mesh size field of the geometric model profile. .

[0041] The inverse distance weighting method with attenuation is as follows: For any point in the non-current sampling point of the geometric model contour... Its relationship with the current sampling point The Euclidean distance is Let i = 1, 2, …, K. The number of current sampling points in the geometric profile is… Select by distance from smallest to largest The closest There are 1 current sampling point, which is a neighboring sampling point, and its number is denoted as . For the numbered neighborhood sampling points, =[1,K], where the grid size contributes to the weight of the size field as follows: The attenuation coefficient is . The mesh size contribution at a certain location serves as the size field of the geometric model profile. The expression is as follows: Equation (6) Equation (7) Equation (8) Attenuation coefficient in this embodiment The value is 2. The value is the greater of the following two: ① The smaller value than 24 (min ( ,24));② The floor value ( ).

[0042] S5-2 Refinement Mesh Generation and Convergence Judgment: Non-uniform mesh size field obtained from S5-2 The Delaunay algorithm is used to construct a mesh, and the S-parameters of the nth round are solved and extracted. If the difference ΔS between the S-parameters of the nth round and the (n-1)th round is less than the threshold τS, then convergence is achieved, and the mesh generated in the (n-1)th round is used for full-band simulation; otherwise, return to step S5-1 to repeat the encryption.

[0043] See attached document Figure 2-3 As shown, the effect of the adaptive mesh generated by the method proposed in this invention can be intuitively observed. Figure 2 The mesh distribution generated based on the initial mesh size is shown. It can be seen that the mesh subdivision is relatively uniform overall, except for some densification at the ports. However, the mesh size is relatively large at the edges and corners of the microstrip lines, which fails to accurately depict the edge effects and current-dense areas during high-frequency signal transmission. Figure 3 The mesh that has reached convergence after being processed by the method of this invention is shown. The mesh density is significantly improved in the edge region of the RF device (i.e., the region with drastic current changes). This is because step S5 of this method accurately identifies and densifies the sampling points with relatively large current changes, and the area from the dense mesh region to the sparse mesh region exhibits a smooth size transition, thanks to the size field generated by the "inverse distance weighting method with attenuation" used in step S5.

[0044] This embodiment also provides an adaptive mesh generation system for electromagnetic field simulation, including: The data acquisition module is responsible for acquiring the geometric model data and electromagnetic characteristic parameters of each radio frequency device model in the integrated circuit; The first calculation module is responsible for extracting the connection relationship information of each radio frequency device based on the geometric model data of each radio frequency device model; merging the geometric model data and the connection relationship information to form a geometric representation containing topological connection information; The second calculation module is responsible for calculating the initial grid size that can initially cover the key area based on electromagnetic property parameters; The third calculation module is responsible for setting current sampling points on the surface of the metal radio frequency device at intervals of a1 times the initial grid size on the non-contact surfaces with other radio frequency devices, based on the geometric representation containing topological connection information; during the simulation, the current value of the current sampling point is obtained by interpolating the vertices of the grid cells where the current sampling point is located. The fourth calculation module is responsible for obtaining the S-parameters and current values ​​of the current sampling points from the first two rounds of mesh generation. If the convergence condition is met, the mesh generated in the first round is used for full-band simulation. If the convergence condition is not met, a local refinement process is performed until the convergence condition is met, and then the mesh generated in the previous round is used for full-band simulation.

[0045] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. An adaptive mesh generation method for electromagnetic field simulation based on electromagnetic field distribution characteristics, characterized in that, The method includes: Obtain the geometric model data and electromagnetic characteristic parameters of each radio frequency device model in the integrated circuit; The connection relationship information of each RF device is extracted from the geometric model data of each RF device model; the geometric model data and the connection relationship information are merged to form a geometric representation containing topological connection information. The initial mesh size that can initially cover the key area is calculated based on electromagnetic property parameters; Based on the geometric representation containing topological connection information, current sampling points are set at intervals of a1 times the initial grid size on the surface of the metal RF device and the non-contact surface with other RF devices, where a1 > 0; during the simulation, the current value of the current sampling point is obtained by interpolating the vertices of the grid cells where the current sampling point is located. The S-parameters and current values ​​at the current sampling points are obtained through the first two rounds of mesh generation. If the convergence condition is met, the mesh generated in the first round is used for full-band simulation. If the convergence condition is not met, a local refinement process is performed until the convergence condition is met, and then the mesh generated in the previous round is used for full-band simulation.

2. The method according to claim 1, characterized in that, The geometric model data includes the contour point coordinates and dimensions of the radio frequency device; the electromagnetic characteristic parameters include the maximum relative permittivity of the metal layer and the maximum simulated frequency.

3. The method according to claim 1, characterized in that, The process of extracting the connection relationship information of each radio frequency device based on the geometric model data of each radio frequency device model is as follows: Generate a 3D axis-aligned bounding box for the geometric model data of each RF device; A three-dimensional KD tree spatial index structure is constructed based on the three-dimensional axis-aligned bounding boxes of all RF devices. Traverse each radio frequency device and use the three-dimensional KD tree spatial index structure to search for a list of potential contact objects that spatially overlap with their bounding boxes; Perform a geometric Boolean intersection operation on the radio frequency devices in the searched list of potential contact objects to obtain the contact surface; The obtained contact surface geometric model data is merged with the initial geometric model data of the RF device to construct a unified data structure that distinguishes between contact surfaces and non-contact surfaces, forming a geometric representation containing complete topological connection information.

4. The method according to claim 1, characterized in that, The process of obtaining S-parameters and current values ​​at current sampling points through the first two rounds of mesh generation; if the convergence condition is met, the mesh generated in the first round is used for full-band simulation; if the convergence condition is not met, local mesh refinement is performed. In the first round, a mesh generation algorithm is used to generate a mesh with a size a2 times the initial mesh size, where a2 > a1. The electromagnetic integral equation at the maximum frequency point of the simulation is solved, and the S-parameters and current values ​​at the current sampling points in the first round are extracted. In the second round, a mesh generation algorithm is used to generate a mesh with a size a3 times the initial mesh size, where a2 > a3 > a1. The electromagnetic integral equation at the maximum frequency point of the simulation is solved in the same way as in the first round, and the S-parameters and current values ​​at the current sampling points in the second round are extracted. If the difference between the two S-parameter values ​​is less than the threshold, convergence is determined, and the first-round mesh is used for full-band simulation; otherwise, the local refinement process is initiated.

5. The method according to claim 1, characterized in that, The implementation process of the local encryption procedure is as follows: (1) Local refinement of coordinate points in the geometric model outline, specifically: In the geometric model contour, coordinate points are divided into current sampling points and non-current sampling points; The grid size is updated by classifying current sampling points into encrypted and unencrypted points; A non-uniform subdivision size field is used to generate the geometric model contour of non-current sampling points; (2) Refinement mesh generation and convergence judgment: Based on the non-uniform partitioned size field, a mesh is constructed using a mesh partitioning algorithm, and the S-parameters of the nth round are solved and extracted; If the difference between the S-parameters in round n and round n-1 is less than the threshold, then convergence is determined, and the mesh generated in round n-1 is used for full-band simulation; otherwise, return to step S5-1 to repeat local refinement.

6. The method according to claim 5, characterized in that, The process of classifying and updating the grid size by using encrypted and unencrypted points for current sampling points is as follows: Calculate the relative rate of change of current at the current sampling points, and select the current sampling points with the largest relative rate of change of current in the top N% as the densification points; if the i-th current sampling point in the nth round is a densification point, its grid size is thinned by the densification coefficient; if the i-th current sampling point in the nth round is a non-densification point, the grid size remains unchanged.

7. The method according to claim 5, characterized in that, The process of generating the non-uniform subdivision size field of the geometric model contour from the non-current sampling points is as follows: Using the grid size of the current sampling points as the control source, a non-uniform subdivision size field of the geometric model profile is generated using an inverse distance weighting method with attenuation.

8. An adaptive mesh generation system for electromagnetic field simulation that implements the method of any one of claims 1-7, characterized in that, include: The data acquisition module is responsible for acquiring the geometric model data and electromagnetic characteristic parameters of each radio frequency device model in the integrated circuit; The first calculation module is responsible for extracting the connection relationship information of each radio frequency device based on the geometric model data of each radio frequency device model; The geometric model data is combined with the connection relationship information to form a geometric representation that includes topological connection information; The second calculation module is responsible for calculating the initial grid size that can initially cover the key area based on electromagnetic property parameters; The third calculation module is responsible for setting current sampling points at intervals of a1 times the initial grid size on the surface of the metal radio frequency device and the non-contact surface with other radio frequency devices, based on the geometric representation containing topological connection information. During the simulation, the current value of the current sampling point is obtained by interpolating the vertices of the grid cell where the current sampling point is located; The fourth calculation module is responsible for obtaining the S-parameters and current values ​​of the current sampling points from the first two rounds of mesh generation. If the convergence condition is met, the mesh generated in the first round is used for full-band simulation. If the convergence condition is not met, a local refinement process is performed until the convergence condition is met, and then the mesh generated in the previous round is used for full-band simulation.

9. A computer-readable storage medium having a computer program stored thereon, which, when executed in a computer, causes the computer to perform the method of any one of claims 1-7.

10. A computing device comprising a memory and a processor, wherein the memory stores executable code, and the processor, when executing the executable code, implements the method of any one of claims 1-7.

Images