Hybrid electromagnetic simulation method suitable for multi-scale fine structure antenna

Abstract

The invention discloses a hybrid electromagnetic simulation method suitable for a multi-scale fine-structure antenna, and the method comprises the steps: firstly creating a multi-scale fine-structure antenna model which needs to be simulated, and determining the range of the size of the multi-scale fine-structure antenna model in a calculation space; constructing a convolution complete matching layer suitable for the hybrid implicit-explicit time domain finite difference method, determining a whole calculation space range including the convolution complete matching layer, then setting a non-uniform grid, and determining a fine region and a transition region from fine to rough; combining the three, adding an excitation source of an electric field, and then obtaining a convolution complete matching layer absorption item, a magnetic field value and an electric field value at the next moment; and finally updating the excitation source in the fine structure direction and updating the time step. According to the method, the hybrid implicit-explicit finite difference time domain method is efficiently combined with the non-uniform grid technology and the convolution complete matching layer, the calculation precision is improved, the implementation process is remarkably simplified, and the calculation efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the field of physics computing, and specifically to electromagnetic simulation of multi-scale fine structures, and proposes a hybrid electromagnetic simulation method suitable for multi-scale fine structure antennas; specifically, it relates to an efficient simulation means combining a hybrid implicit-explicit time-domain finite-difference method, a non-uniform grid technology and a convolutional perfectly matched layer in a Cartesian coordinate system. Background Art

[0002] Absorption boundaries are designed to address the challenges encountered when simulating electromagnetic propagation problems in infinite space in computational electromagnetics. In electromagnetic problems, electromagnetic waves usually propagate in infinite space. However, in numerical simulations, the computational domain must be finite. If no measures are taken, electromagnetic waves will be reflected back at the boundary of the computational domain, resulting in inaccurate calculation results. At present, the most representative absorption boundaries are Mur absorption boundaries, perfectly matched layers (PML) and convolutional perfectly matched layers (CPML). CPML is an improved version of PML, which can provide better numerical stability and better absorption effect than traditional PML.

[0003] The development of non-uniform grid technology began with the application requirements of electromagnetic solution methods in high-frequency electromagnetic wave propagation, complex geometry simulation, and various inhomogeneous media. Using uniform grids to simulate electromagnetic fields means that all points in the entire simulation area have the same spatial step size. This may be sufficient in simple geometric structures, but when dealing with problems with complex geometries or multi-size features, uniform grids can lead to huge computational burdens. Non-uniform grid technology allows the use of finer grids in these critical areas, while maintaining coarser grids in other areas, thereby improving the accuracy and stability of the calculation.

[0004] Nowadays, computer technology is developing at a high speed, and electromagnetic simulation problems are becoming more and more complicated. In this case, traditional analytical methods are difficult to meet the needs of complex electromagnetic simulation due to their own limitations, so numerical calculation methods have taken their place and become the mainstream means to solve such problems. Among them, the Finite-Difference Time-Domain (FDTD) method, as one of the three major mainstream numerical algorithms, has significant advantages and is widely used in many research fields. It has become an indispensable key method in this field and provides solid support for cutting-edge exploration and technical breakthroughs. However, due to the limitations of the time step of the FDTD method and low computational efficiency, the traditional FDTD method cannot efficiently and accurately simulate and analyze fine structures. In response to these shortcomings, experts have made improvements and proposed the Hybrid Implicit-Explicit Finite-Difference Time-Domain (HIE-FDTD), which reduces the number of iterations required in the computational domain in terms of computational efficiency, thereby improving the overall computational efficiency. At the same time, by introducing the implicit method, the HIE-FDTD method can use a larger time step without strictly following the time step limit, thereby improving the stability of the time step. However, facing the dilemma that this method cannot be effectively applied to the far-field radiation problem of microwave devices with multi-scale fine structures, a grid generation method suitable for the HIE-FDTD method and an efficient and stable absorbing boundary are urgently needed. Summary of the invention

[0005] In order to overcome the defects of the prior art, the purpose of the present invention is to provide an efficient hybrid grid and boundary electromagnetic adaptation method for multi-scale fine structures, and propose a method based on a hybrid implicit-explicit time-domain finite difference method combined with non-uniform grid technology and convolutional perfectly matched layer in a Cartesian rectangular coordinate system. This method can effectively reduce the reflection of electromagnetic waves at the boundary by combining non-uniform grid technology and convolutional perfectly matched layer. At the same time, this method can more accurately capture the subtle changes of the electromagnetic field in key parts, and can also better deal with numerical stability problems under high absorption conditions, reducing numerical oscillations and errors.

[0006] The technical solution adopted by the present invention to solve the technical problem is:

[0007] In a first aspect, the present invention provides a hybrid electromagnetic simulation method applicable to a multi-scale fine structure antenna, wherein the multi-scale fine structure refers to a multi-size fine structure in only one spatial direction; the method comprises the following steps:

[0008] Step S1, creating a multi-scale fine structure antenna model to be simulated, and determining the range of its size in the calculation space;

[0009] Step S2, constructing a convolutional perfectly matched layer suitable for a hybrid implicit-explicit finite-difference time-domain method, and determining the entire calculation space range including the convolutional perfectly matched layer, setting a non-uniform grid for a multi-scale fine structure antenna model to be simulated, and determining a fine area and a non-uniform area from fine to coarse;

[0010] The fine area adopts a high-resolution fine grid, and the coarse area adopts a low-resolution coarse grid;

[0011] Step S3, constructing a hybrid algorithm: introducing a convolutional perfectly matched layer absorbing boundary condition into the hybrid implicit-explicit time-domain finite difference method to obtain the implicit equation part; then combining the implicit equation part with the non-uniform grid, redefining the difference operator according to the grid distribution, so that the difference operator adapts to the non-uniform step size;

[0012] Step S4, adding an excitation source in the direction of the fine structure, and obtaining the convolution perfectly matched layer absorption term, magnetic field value, and electric field value at the next moment based on the hybrid algorithm constructed in step S3;

[0013] Step S5, updating the excitation source in the fine structure direction and updating the time step t;

[0014] Assume that the range of time step t is 1 to T, and the time step t increases by 1 in each iteration; determine whether the time step t reaches T, if not, the time step t increases by 1, and returns to step S4 to continue iteration; if it reaches (that is, when t = T), the iteration ends, and the magnetic field value and electric field value obtained in step S4 for each iteration are recorded.

[0015] Preferably, the multi-scale fine structure is a dual-frequency microstrip patch antenna with a multi-scale fine structure.

[0016] In a second aspect, the present invention provides an electromagnetic simulation system for implementing the above method, comprising the following modules:

[0017] Model creation module, used to create multi-scale fine structure antenna models that need to be simulated;

[0018] A hybrid module for combining the hybrid implicit-explicit finite-difference time-domain method, convolutional perfectly matched layers and non-uniform grids to build a hybrid algorithm;

[0019] The simulation module is used to simulate multi-scale fine structure antenna models using a hybrid algorithm.

[0020] In a third aspect, 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 execute the above method.

[0021] In a fourth aspect, the present invention provides a computing device, comprising a memory and a processor, wherein the memory stores executable code, and when the processor executes the executable code, the above method is implemented.

[0022] Compared with the prior art, the present invention has the following advantages:

[0023] The present invention introduces an efficient combination strategy of hybrid implicit-explicit finite-difference time-domain method, non-uniform grid technology and convolutional perfectly matched layer. For far-field simulation with multi-scale fine structures, finer grids are used in areas with drastic changes in electromagnetic fields or complex structures, while coarser grids are used in areas with gentler changes. This strategy can significantly reduce the number of grid points that need to be calculated, simplify the implementation process, thereby reducing the overall amount of calculation and improving calculation efficiency. In addition, non-uniform grids allow more accurate simulation of important detail areas in electromagnetic simulation. For example, in key areas such as slits, tips, and antenna feed points in microwave circuits, higher resolution grids can be used to accurately capture subtle changes in the electromagnetic field. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 It is an algorithm flow chart of the implementation method of the present invention.

[0025] Figure 2 It is a schematic diagram of the geometric model of the antenna structure of an embodiment of the implementation method of the present invention; wherein (a) is the front side of the antenna in embodiment 1 of the present invention, and (b) is the back side of the antenna.

[0026] Figure 3 It is an S parameter diagram of the antenna in Example 1 of the present invention compared with a high-frequency simulator under different CFLN (Courant-Friedrich-Levy number, stability constant) conditions.

[0027] Figure 4 It is an S parameter diagram of the antenna in Example 1 of the present invention compared with a high frequency simulator under different CFLN conditions.

[0028] Figure 5 1 is the gain radiation pattern of the antenna in Example 1 of the present invention under different CFLN conditions; wherein (a) is a comparison curve diagram at 2.4 GHz under different azimuth angles phi; (b) is a comparison curve diagram at 3.7 GHz under different azimuth angles phi; (c) is a comparison curve diagram at 2.4 GHz under different azimuth angles theta; (d) is a comparison curve diagram at 3.7 GHz under different azimuth angles theta. DETAILED DESCRIPTION

[0029] The present invention is further described below in conjunction with embodiments and drawings.

[0030] The present invention is applied to the hybrid implicit-explicit time-domain finite-difference method in computational electromagnetics. The recursive formula involved can be obtained according to the hybrid implicit-explicit time-domain finite-difference method. The specific implementation method is described by taking the electromagnetic simulation problem of the antenna as an example. The actual application scope is not limited to this, and is also applicable to the fields of wireless communications, integrated circuits, etc.

[0031] The present invention provides a method for implementing a hybrid non-uniform grid technology and a convolutional perfect matching layer, specifically a method based on a hybrid implicit-explicit finite difference time-domain method combined with a non-uniform grid technology and a convolutional perfect matching layer in a Cartesian rectangular coordinate system. Figure 1 As shown, the specific steps of the method are:

[0032] (1) Create a multi-scale fine structure model that needs to be simulated and determine the range of its size in the computational space.

[0033] First, a multi-scale fine structure simulation model is created, and then the size range of the antenna is determined based on the model. The size range of the antenna is defined as (X min ,Y min ,Z min ) to (X max ,Y max ,Z max ). Among them, X min and X max Represents the minimum and maximum positions on the x-axis, Y min and Y max Represents the minimum and maximum positions on the y-axis, Z min and Z max Represents the minimum and maximum positions on the z-axis respectively.

[0034] (2) Construct a convolutional perfectly matched layer suitable for the hybrid implicit-explicit finite-difference time-domain method and determine the entire computational space range including the convolutional perfectly matched layer. For the multi-scale fine structure antenna model that needs to be simulated, set a non-uniform grid and determine the fine area and the non-uniform area from fine to coarse.

[0035] Without loss of generality, the three-dimensional Maxwell equations inside the convolution perfectly matched layer in a lossy medium are expressed as:

[0036]

[0037]

[0038]

[0039]

[0040]

[0041]

[0042] Among them, j represents an imaginary number, ω represents the angular velocity, ε η (η=x,y,z) represents the dielectric constant in the η direction, represents the spatial difference operator of the first-order derivative along the η direction, μ η represents the magnetic permeability in the η direction, E η represents the electric field in the η direction, H η represents the magnetic field in the η direction, and is the conductivity in the cut-off medium, S eη and S mη Represents the stretching coordinate factor, see formula (7) for details:

[0043]

[0044] Among them, ε0 represents the vacuum dielectric constant, σ eη , σ mη , κ η , α eη and α mη are all non-negative real numbers, expressed as:

[0045]

[0046]

[0047]

[0048] Among them, σ eη and σ mη represents the conductivity, σ ηmax Represents σ η The maximum value of , ρ represents the distance from the electromagnetic field to the interface between the calculation area and the absorption layer, d represents the thickness of the absorption layer, m represents the order of the polynomial, Δη represents the spatial step in the η direction, κ ηmax Represents κ η The maximum value of α min and α max Represents α η The minimum and maximum values ​​of

[0049] According to the antenna size range and the absorption layer thickness d determined by the model, the size range of the calculation space is defined as (X min -d,Y min -d,Z min -d) to (X max +d,Y max +d,Z max+d). Among them, X min -d and X max +d represent the minimum and maximum positions of the calculation space on the x-axis, Y min -d and Y max +d represent the minimum and maximum positions of the calculation space on the y-axis, Z min -d and Z max +d represent the minimum and maximum positions of the calculation space on the z-axis.

[0050] The non-uniform grid technology is divided into fine area and non-uniform area. The fine area adopts fine grid, and the non-uniform area adopts non-uniform grid. First, the direction parameter dir='η' of the non-uniform grid technology is set. Then, the size of the fine area and the size of the fine grid in the fine area are set. The size of the fine area is rΔs, where Δs represents the size of the fine grid in the fine area, and r represents the number of fine grids. Finally, the size of the grid in the non-uniform area is determined. The size of the grid in the non-uniform area is determined by ΔM, which can be expressed as:

[0051] ΔM=R M Δs (11)

[0052] Among them, R represents the rate of change between subsequent half grids, and M represents the index of the grid in the non-uniform area. The grid size of the coarse area adjacent to the non-uniform area is expressed as:

[0053]

[0054] Among them, N nu represents the number of grids in the non-uniform region. Therefore, the total length of the non-uniform region plus one grid at both ends of the uniform region can be expressed as:

[0055]

[0056] R, N nu It can be obtained from formula (13):

[0057]

[0058] (3) The hybrid implicit-explicit finite-difference time-domain method is effectively combined with the convolutional perfectly matched layer to obtain the implicit equation part of the hybrid implicit-explicit finite-difference time-domain method combined with the convolutional perfectly matched layer.

[0059] Without loss of generality, the fine structure is defined as the z direction. According to the hybrid implicit-explicit FDTD method, the three-dimensional hybrid implicit-explicit FDTD method with convolutional perfectly matched layer is expressed as:

[0060]

[0061]

[0062]

[0063]

[0064]

[0065]

[0066] Where n, n+1 / 2 and n+1 represent time indices, Δt is the time step, is an auxiliary function; a1=Δt / ε, b1=Δt / μ, ε represents the dielectric constant, μ represents the magnetic permeability; represents Θ at time T1 η , Θ=E, H, T1=n-1 / 2,n,n+1 / 2,n+1; and They represent the electric field E in the η2 direction at time T2. η1 and magnetic field H η1 The corresponding absorption term, T2 = n,n + 1 / 2,

[0067] Substituting formula (20) into formula (17) to eliminate We can get:

[0068]

[0069] Substituting formula (19) into formula (18) to eliminate We can get:

[0070]

[0071] The absorption term It can be expressed as:

[0072]

[0073]

[0074]

[0075]

[0076]

[0077]

[0078]

[0079]

[0080]

[0081]

[0082]

[0083]

[0084] where c eη 、c mη d eη and d mη are all non-negative real numbers and can be expressed as:

[0085]

[0086]

[0087] (4) The equations after the hybrid implicit-explicit finite-difference time-domain method is effectively combined with the convolutional perfectly matched layer, and then combined with the non-uniform grid technology.

[0088] The spatial step length of the non-uniform grid in three directions is defined as Δx i ,Δy j ,Δz k Δx i represents the non-uniform grid size in the x direction, Δy j represents the non-uniform grid size in the y direction, Δz k represents the non-uniform grid size in the z direction. η Re-expressed as Δη is replaced by a non-uniform grid Δx i ,Δy j ,Δz k :

[0089]

[0090]

[0091]

[0092]

[0093]

[0094]

[0095] (5) Add an excitation source in the direction of the fine structure, and obtain the convolutional perfectly matched layer absorption term, magnetic field value, and electric field value at the next moment based on an algorithm combining a hybrid implicit-explicit finite-difference time-domain method with a convolutional perfectly matched layer and a non-uniform grid technology;

[0096] Adding an excitation source J in the z direction of the fine structure, that is, adding the required excitation source to equation (37), is expressed as:

[0097]

[0098] According to the algorithm combining the hybrid implicit-explicit finite-difference time-domain method with the convolutional perfectly matched layer and the non-uniform grid technology, namely, equations (38) to (43), the convolutional perfectly matched layer absorption term, magnetic field value, and electric field value at the next moment are obtained through iteration.

[0099] (6) Update the excitation source in the fine structure direction and update the time step t;

[0100] The time step t will iterate in a loop, and the range of t is between 1 and T. After each loop, t will increase by 1, and at the same time, it will be checked whether it has reached T. If the time step t has not reached T, it will return to step S5 again. When t is equal to T, the number of iteration steps is reached, and the loop ends. The magnetic field and electric field values ​​obtained in step S5 for each loop are recorded.

[0101] Example 1

[0102] This embodiment is an example of electromagnetic simulation of an antenna, and all simulation results are derived from the implementation method of the present invention.

[0103] See also Figure 2 , is a schematic diagram of the geometric model of the antenna structure used in this embodiment, the thickness of the dielectric substrate is 1.6 mm, and the other geometric dimension parameters are detailed in Table 1:

[0104] Table 1 Geometric dimensions of the antenna in Example 1

[0105] parameter W <![CDATA[W1]]> <![CDATA[W2]]> <![CDATA[W3]]> L <![CDATA[L1]]> <![CDATA[L2]]> <![CDATA[L3]]> <![CDATA[L4]]> <![CDATA[L5]]> Dimensions (mm) 38.00 16.15 4.15 2.50 19.00 4.00 9.50 0.50 5.50 4.00

[0106] The implementation process of the embodiment is the same as that of the specific implementation method; the Gaussian excitation source J is set at the z-direction center position of the side surface in the fine structure direction in the simulation space, and the spatial step size of the fine grid area is Δx=0.35mm, Δy=0.50mm, Δz=0.04mm. The spatial step size of the coarse grid area is Δx=Δy=0.50mm, Δz=0.04mm. The fine grid spatial step size of the non-uniform grid area is Δx=0.35mm, Δy=0.50mm, Δz=0.04mm, and the coarse grid is Δx=0.50mm. The grid resolution is 20, CFLN is 1, 3, and 5 respectively, and the convolutional fully matched layer thickness is 8 layers in the x-, y-, and z-directions.

[0107] The embodiment 1 was measured in a computer. It should be noted that all the measurement results of the method of the present invention are measured in the same measurement environment. The computer CPU is i5-14600K and the running memory is 32G.

[0108] The results are as follows Figure 3 , Figure 4 shown. Figure 3 , Figure 4 represents the time step ratio between the HIE-FDTD method involved in the method of the present invention and the traditional FDTD method, and the CFLN values ​​are 1, 3, and 5 respectively; wherein, Figure 3 and Figure 4 The methods involved include a combination of the FDTD method and CPML (i.e., the FDTD-CPML method), a combination of the HIE-FDTD method and CPML using fine grid simulation (i.e., the FG-HIE-CPML method, referred to as FG in the figure), a combination of the HIE-FDTD method with non-uniform grid technology and CPML (i.e., the NG-HIE-CPML method, referred to as NG in the figure), and a combination of the HIE-FDTD method and CPML using coarse grid simulation (i.e., the CG-HIE-CPML method, referred to as CG in the figure).

[0109] Specifically, Figure 3 The S parameters and simulation results obtained by the method of the present invention and the FG-HIE-CPML method are shown in the figure. Even if the CFLN is increased to 5, the simulation results of the method of the present invention are still very consistent with the results of the High Frequency Structural Simulator (HFSS) and the FDTD-CPML method, which effectively verifies its high calculation accuracy. Among them, the method of the present invention is a combination of the HIE-FDTD method with the non-uniform grid technology (NG) and CPML (i.e., the NG-HIE-CPML method).

[0110] Specifically, Figure 4 The S parameters and simulation results obtained by the method of the present invention and the CG-HIE-CPML method are shown in Figure 2. By comparing the method of the present invention with other methods, the simulation results of the method of the present invention are very consistent with the results of the FDTD-CPML method, which once again verifies the high calculation accuracy of the HIE-CPML method using non-uniform grid technology.

[0111] Figure 5The simulation results of the radiation pattern of the antenna at 2.4GHz and 3.7GHz are shown, showing the gain distribution on the xoy and xoz planes, where CFLN takes values ​​of 1, 3, and 5 respectively. Dphi in the figure represents the change of the antenna's directivity coefficient at different azimuth angles phi, and Dtheta represents the change of the antenna's directivity coefficient at different azimuth angles theta. Ph and theta are parameters used to describe the radiation directivity of the antenna. It can be observed that although the simulation results of the NG-HIE-CPML method have some deviations with the increase of CFLN, they still maintain a high consistency overall, and have little effect on the overall simulation and evaluation of the antenna performance, which effectively verifies the high stability of the method of the present invention. When CFLN rises to 5, it can still be seen that the difference with the results of FDTD-CPML is very small, which reflects the accuracy of the NG-HIE-CPML method and verifies the effective combination of non-uniform grid technology, CPML and HIE-FDTD method.

[0112] The comparison of the calculation results of different algorithms is shown in Table 2. As CFLN increases, it can be seen that the time required for NG-HIE-CPML is shortened. When CFLN rises to 5, the calculation time of the method of the present invention is 46.66% faster than that of the traditional FDTD-CPML method, which proves its high calculation efficiency.

[0113] Table 2 Calculation results of different methods in Example 1

[0114]

[0115] In summary, the method of the present invention simulates a dual-band microstrip patch antenna and compares the simulation results of the S parameters and radiation patterns of the method of the present invention with those of the HFSS software, thereby verifying its accuracy and effectiveness in processing the simulation of multi-scale fine structure models; compared with the traditional FDTD-CPML method, the calculation time of the method of the present invention is only 53.34% of that of the traditional FDTD-CPML method, which proves its high calculation efficiency.

[0116] Although the present invention has been disclosed as above with preferred embodiments, it is not intended to limit the present invention. A person with ordinary knowledge in the technical field to which the present invention belongs may make various changes and modifications without departing from the spirit and scope of the present invention. Therefore, the protection scope of the present invention shall be determined by the definition of the claims.

Claims

1. A hybrid electromagnetic simulation method suitable for multi-scale fine structure antennas, characterized in that: The method comprises the following steps: Step S1, creating a multi-scale fine structure antenna model to be simulated, and determining the range of its size in the calculation space; Step S2, constructing a convolutional perfectly matched layer suitable for a hybrid implicit-explicit finite-difference time-domain method, and determining the entire calculation space range including the convolutional perfectly matched layer, setting a non-uniform grid for a multi-scale fine structure antenna model to be simulated, and determining a fine area and a non-uniform area from fine to coarse; Step S3, constructing a hybrid algorithm: introducing a convolutional perfectly matched layer absorbing boundary condition into the hybrid implicit-explicit time-domain finite difference method to obtain the implicit equation part; then combining the implicit equation part with the non-uniform grid, redefining the difference operator according to the grid distribution, so that the difference operator adapts to the non-uniform step size; Step S4, adding an excitation source in the direction of the fine structure, and obtaining the convolution perfectly matched layer absorption term, magnetic field value, and electric field value at the next moment based on the hybrid algorithm constructed in step S3; Step S5, update the excitation source in the fine structure direction and update the time step t; set the range of the time step t to 1 to T, and the time step t increases by 1 in each iteration; determine whether the time step t reaches T, if not, increase the time step t by 1, and return to step S4 to continue the iteration; if reached, the iteration ends, and the magnetic field value and electric field value obtained in step S4 for each iteration are recorded.

2. The method according to claim 1, characterized in that: The size range of the multi-scale fine structure antenna model in step S1 is defined as (X min ,Y min ,Z min ) to (X max ,Y max ,Z max ), where X min and X max Represents the minimum and maximum positions on the x-axis, Y min and Y max Represents the minimum and maximum positions on the y-axis, Z min and Z max Represents the minimum and maximum positions on the z-axis respectively.

3. The method according to claim 1, characterized in that: In step S2, the three-dimensional Maxwell equations inside the convolution perfectly matched layer in the lossy medium are expressed as: Where η = x, y, z; j represents an imaginary number, ω represents the angular velocity, ε η represents the dielectric constant in the η direction, represents the spatial difference operator of the first-order derivative along the η direction, μ η Indicates the magnetic permeability in the η direction, E η represents the electric field in the η direction, H η represents the magnetic field in the η direction, and is the conductivity in the cut-off medium, S eη and S mη Represents the stretching coordinate factor, see formula (7) for details: Where ε0 represents the vacuum dielectric constant, σ eη , σ mη , κ η , α eη and α mη are all non-negative real numbers, expressed as: Among them, σ eη and σ mη represents the conductivity, σ ηmax Represents σ η The maximum value of , ρ represents the distance from the electromagnetic field to the interface between the calculation area and the absorption layer, d represents the thickness of the absorption layer, m represents the order of the polynomial, Δη represents the spatial step in the η direction, κ ηmax Represents κ η The maximum value of α min and α max Represents α η The minimum and maximum values ​​of According to the antenna size range and the absorption layer thickness d determined by the model, the size range of the calculation space is defined as (X min -d,Y min -d,Z min -d) to (X max +d,Y max +d,Z max +d).

4. The method according to claim 1, characterized in that: In step S2, firstly, the direction parameter dir='η' of the non-uniform grid is set, then the fine region size and the fine region fine grid size are set, the fine region size is nΔs, Δs represents the fine region fine grid size, n represents the number of fine grids, and finally the non-uniform region grid size is determined, the non-uniform region grid size is determined by ΔM, expressed as: ΔM=R M Δs (11) Where R represents the rate of change between subsequent half-grids, and M represents the index of the grid in the non-uniform region; The mesh size of the coarse area adjacent to the non-uniform area is expressed as: Among them, N nu Indicates the number of grids in the non-uniform area; Therefore, the total length of the non-uniform region plus one grid at each end of the uniform region is expressed as: From formula (13), we can get R and N nu :

5. The method according to claim 1, characterized in that: In step S3, the implicit equation part is obtained by the following method: The fine structure direction is defined as the z direction. According to the hybrid implicit-explicit FDTD method, the three-dimensional hybrid implicit-explicit FDTD method with convolutional perfectly matched layers is expressed as: Where n, n+1 / 2 and n+1 represent time indices, Δt is the time step, is an auxiliary function; a1=Δt / ε, b1=Δt / μ, ε represents the dielectric constant, μ represents the magnetic permeability; represents Θ at time T1 η , Θ=E, H, T1=n-1 / 2,n,n+1 / 2,n+1; and They represent the electric field E in the η2 direction at time T2. η1 and magnetic field H η1 The corresponding absorption term, T2 = n,n + 1 / 2, Substituting equation (20) into equation (17) to eliminate H y n+1 We can get: Substituting equation (19) into equation (18) to eliminate H x n+1 We can get: The absorption term It can be expressed as: where c eη 、c mη d eη and d mη are all non-negative real numbers and can be expressed as:

6. The method according to claim 5, characterized in that: In step S3, the implicit equation part is combined with the non-uniform grid as follows: The spatial step length of the non-uniform grid in three directions is defined as Δx i ,Δy j ,Δz k , where Δx i represents the non-uniform grid size in the x direction, Δy j represents the non-uniform grid size in the y direction, Δz k represents the non-uniform grid size in the z direction; replace δ in equations (15), (16), (19) to (22) η Re-expressed as Δη is replaced by a non-uniform grid Δx i ,Δy j ,Δz k :

7. The method according to claim 6, characterized in that: The specific process of step S4 is: Adding the required excitation source J in equation (37) yields: According to equations (38) to (43), the convolution perfectly matched layer absorption term, magnetic field value, and electric field value at the next moment are obtained through iteration.

8. An electromagnetic simulation system for implementing the method according to any one of claims 1 to 7, characterized in that: Includes the following modules: Model creation module, used to create multi-scale fine structure antenna models that need to be simulated; A hybrid module for combining the hybrid implicit-explicit finite-difference time-domain method, convolutional perfectly matched layers and non-uniform grids to build a hybrid algorithm; The simulation module is used to simulate multi-scale fine structure antenna models using a hybrid algorithm.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed in a computer, the computer is caused to execute the method according to any one of claims 1 to 7.

10. A computing device comprising a memory and a processor, characterized in that: The memory stores executable code, and when the processor executes the executable code, the method according to any one of claims 1 to 7 is implemented.

Images