Design optimization method of pixelated electromagnetic structure, electromagnetic structure and equipment

Abstract

The invention relates to the technical field of electromagnetic structures, in particular to a design optimization method of a pixelated electromagnetic structure, the electromagnetic structure and equipment, and the method comprises the steps: obtaining a target electromagnetic parameter of a target electromagnetic structure, and generating an initial population according to the target electromagnetic parameter; performing performance evaluation processing on the initial population to obtain an error value of the initial population; if the error value of the initial population is not in the target error range, micro optimization processing and macro optimization processing are carried out on the initial population to obtain a current population, and performance evaluation processing is carried out on the current population to realize iterative processing of the current population; determining the current population as a target population until the error value of the current population is within a target error range or the number of iteration processing times of the current population reaches a preset number of times; and screening out a target individual from the target population, and determining the target individual as a target electromagnetic structure. According to the method, the design optimization efficiency of the electromagnetic structure is improved, and the design optimization universality of the electromagnetic structure is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of electromagnetic structure, and particularly relates to a design optimization method of pixelated electromagnetic structure, an electromagnetic structure and equipment. BACKGROUND

[0002] In the design of radio frequency, microwave and millimeter wave circuits and systems, the performance of the radio frequency system is highly dependent on the design of the matching network between the ports and the modules. A good matching network can minimize signal reflection, reduce transmission loss, and improve the stability and reliability of the entire radio frequency system. At present, since the differential circuit has stronger anti-interference ability than the single-ended circuit, the differential circuit has become the mainstream circuit topology form of the radio frequency system. In addition, electromagnetic structures such as coupling transformers are widely used in the matching network of the differential circuit due to their natural differential form and impedance transformation capability. Electromagnetic structures such as coupling transformers have become important elements in realizing the matching network of high-performance radio frequency differential systems, which ingeniously combines circuit functions and performance requirements together, and makes an important contribution to the high performance of modern wireless communication systems.

[0003] However, at present, the design method of high-performance electromagnetic structures such as coupling transformers still adopts the design method of coupling transformers with conventional topologies: the matching network design of the traditional transformer relies on the experience of the designer, and the designer manually adjusts a set of pre-defined geometric parameters (such as wire width, spacing, etc.) and performs electromagnetic simulation verification to design a coupling transformer with a conventional topology. Although optimization algorithms are introduced later to automate this design process, the essence of this design process is still to search for the optimal parameters within a fixed topology structure. The disadvantage of this method is that parameterized modeling is required for each specific transformer topology, which leads to tedious parameterized modeling, wasting time and labor costs. Moreover, the design space of this method is limited to the pre-set topology structure, making it difficult to find more optimal non-traditional topology structures. Therefore, the existing design method of electromagnetic structures such as coupling transformers has the problem of low design optimization efficiency. SUMMARY

[0004] The embodiments of the present application provide a design optimization method of pixelated electromagnetic structure, an electromagnetic structure and equipment, which solves the technical problem of low design optimization efficiency of the existing design method of electromagnetic structures such as coupling transformers, realizes electromagnetic structures through a pixelated optimization design process, and further realizes the technical effects of utilizing the high-dimensional and high-degree design space brought by pixelation, ensuring the generation of basic conductive paths with physical meaning in the initial start-up stage of optimizing electromagnetic structures, thereby greatly improving the design optimization efficiency, reliability and practicality of electromagnetic structures while ensuring the upper limit of performance, and improving the universality and generality of the design optimization of electromagnetic structures.

[0005] In a first aspect, the embodiments of the present application provide a design optimization method of a pixelated electromagnetic structure, comprising: obtaining design parameters of a target electromagnetic structure, the design parameters comprising: process parameters of metal layers in the target electromagnetic structure, a pixel grid map of the metal layers, and a target electromagnetic parameter of the target electromagnetic structure; generating an initial population according to the target electromagnetic parameter, wherein each individual in the initial population carries a key point list, and each key point in the key point list of each individual is a key pixel point on the pixel grid map of each individual; performing performance evaluation processing on the initial population to obtain an error value of the initial population; if the error value of the initial population is not located in a target error range, performing micro-optimization processing and macro-optimization processing on the initial population based on the key point list of each individual in the initial population to obtain a current population, and performing performance evaluation processing on the current population to realize iterative processing of the current population until the error value of the current population is located in the target error range or the number of iterations of the current population reaches a preset number, and determining the current population as a target population; selecting a target individual from the target population, and determining the target individual as the target electromagnetic structure.

[0006] Based on the same inventive concept, in a second aspect, the present application further provides an electromagnetic structure formed by the design optimization method of the pixelated electromagnetic structure of the first aspect.

[0007] Based on the same inventive concept, in a third aspect, the present application provides an electronic device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the processor implements the steps of the design optimization method of the pixelated electromagnetic structure when executing the program.

[0008] The one or more technical solutions in the embodiments of the present application have at least the following technical effects or advantages: Through the design optimization method of the pixelated electromagnetic structure, the optimization design process of pixelation is realized to realize the target electromagnetic structure of multiple layers. In this way, the high-dimensional and high-degree design space brought by pixelation can be utilized, and the basic conductive path with physical meaning can be ensured to be generated in the initial start stage of optimizing the electromagnetic structure, the precise control of complex electromagnetic performance is realized, and the performance upper limit of the electromagnetic structure is ensured. At the same time, the design optimization efficiency, reliability and practicability of the electromagnetic structure are greatly improved, and the universality and generality of the design optimization of the electromagnetic structure are improved. BRIEF DESCRIPTION OF DRAWINGS

[0009] Various other advantages and benefits will become apparent to those of ordinary skill in the art upon reading the following detailed description of the preferred embodiments with reference made to the accompanying drawings. The drawings are for purposes of illustration only and are not intended to be limiting in Figure 1 A flow chart of the steps of the design optimization method of the pixelated electromagnetic structure in the embodiment of the present application is shown; Figure 2 Another flow chart of the steps of the design optimization method of the pixelated electromagnetic structure in the embodiment of the present application is shown; Figure 3 A structure diagram of the pixel grid of the high metal layer and the low metal layer in the embodiment of the present application is shown; Figure 4 A structure diagram of the pixel architecture of an individual in the embodiment of the present application is shown; Figure 5 A pixel encoding diagram of the high metal layer of the individual in the embodiment of the present application is shown; Figure 4 Figure 6a A structure diagram of the optimization process of the comparison result one of the local area around the key point P4 of the pixel encoding matrix of the high metal layer of the individual in the embodiment of the present application is shown; Figure 4 Another structure diagram of the optimization process of the comparison result one of the local area around the key point P4 of the pixel encoding matrix of the high metal layer of the individual in the embodiment of the present application is shown; Figure 6b Figure 4 A structure diagram of the optimization process of the comparison result three of the local area around the key point P4 of the pixel encoding matrix of the high metal layer of the individual in the embodiment of the present application is shown; Figure 7 Another structure diagram of the optimization process of the comparison result three of the local area around the key point P4 of the pixel encoding matrix of the high metal layer of the individual in the embodiment of the present application is shown; Figure 4 A structure diagram of the internal cross processing of another individual in the embodiment of the present application is shown; Figure 8 A structure diagram of the key point addition processing of the high metal layer of the individual shown in the embodiment of the present application is shown; Figure 9 Figure 8 A structure diagram of the key point deletion processing of the high metal layer of the individual shown in the embodiment of the present application is shown; Figure 10 A structure diagram of the key point movement processing of the high metal layer of the individual shown in the embodiment of the present application is shown. Figure 8 Figure 11 Figure 8 DETAILED DESCRIPTION ​​​​​​

[0010] Exemplary embodiments of the present disclosure will be described in greater detail below with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it is understood that the present disclosure can be embodied in various forms and should not be limited by the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the present disclosure to those skilled in the art.

[0011] Embodiment one The first embodiment of the present application provides a design optimization method of a pixelated electromagnetic structure, as shown in Figure 1 and Figure 2 , comprising: S101, obtaining the design parameters of the target electromagnetic structure, the design parameters including: the process parameters of the metal layer in the plurality of metal layers of the target electromagnetic structure and the pixel grid map of the metal layer, and the target electromagnetic parameter of the target electromagnetic structure; S102, generating an initial population according to the target electromagnetic parameter, wherein each individual in the initial population carries a key point list, and the key points in the key point list of each individual are key pixel points on the pixel grid map of each individual; S103, performing performance evaluation processing on the initial population to obtain the error value of the initial population; S104, if the error value of the initial population is not located in the target error range, performing microscopic optimization processing and macroscopic optimization processing on the initial population based on the key point list of each individual in the initial population to obtain a current population, and then performing performance evaluation processing on the current population to realize iterative processing of the current population, until the error value of the current population is located in the target error range or the number of iterations of the current population reaches a preset number, and the current population is determined as a target population; S105, selecting a target individual from the target population, and determining the target individual as the target electromagnetic structure.

[0012] It should be explained that the population is a collection of individuals, and an individual is a specific structure of an electromagnetic structure. The initial population is the population of the first iteration processing, and the initial population is also the current population in the first iteration processing. The next population of the initial population is the population of the second iteration processing, and the next population is also the current population in the second iteration processing. The population in each iteration processing is called the current population. The iteration processing process of the population is described in detail in this embodiment taking the initial population as an example, and the iteration processing process of other populations (i.e. the current population) is consistent with that of the initial population.

[0013] In the embodiment, after the design parameters of the target electromagnetic structure are obtained, the initial population is generated based on the target electromagnetic parameters of the target electromagnetic structure. In this way, a plurality of individuals are randomly generated according to the target electromagnetic parameters of the target electromagnetic structure, and the individuals are completely different, providing a rich sample of individuals that meet the target electromagnetic parameters and improving the diversity and automation of the design optimization process of the target electromagnetic structure, reducing the limitations of the design optimization process, and improving the design optimization efficiency and practicality. Moreover, each individual in the initial population carries a key point list, and the key points in the key point list of each individual are key pixel points on the pixel grid map of each individual. The key point list is a coherent metal scatter point of the individual, and based on the key point list, a conductive path of the individual can be established, so that the individual forms a complete structure of effective electromagnetic coupling. Through the key point list of each individual, the physical connectivity of the individual structure is ensured, providing a solid foundation for the design optimization process of the population and individuals, and improving the design optimization efficiency of the electromagnetic structure.

[0014] After obtaining the initial population, the performance evaluation process is performed on the initial population to obtain the error value of the initial population. In this way, it is determined whether there is a target individual, i.e., a target electromagnetic structure, in the initial population based on the error value of the initial population. In this way, it can be quickly and simply determined whether the population contains a target electromagnetic structure, improving the design optimization efficiency. When it is determined that the error value of the initial population is not located in the target error range, it indicates that there is no target individual, i.e., a multi-layer target electromagnetic structure, in the initial population, and the initial population needs to be iteratively processed to obtain a current population until the error value of the current population is located in the target error range or the number of iterations of the current population reaches a preset number, and the current population is determined as a target population. The target individual is selected from the target population, and the target individual is determined as the target electromagnetic structure.

[0015] The iterative processing process is that the initial population is micro-optimized and macro-optimized based on the key point list carried by each individual in the initial population to obtain a current population, and the performance evaluation process is performed on the current population. The iterative processing process is based on the optimization genes formed by the key point list of the individuals in the population to macro-optimize and micro-optimize the individuals in the population, i.e., to globally optimize and locally optimize the individuals, so that the actual electromagnetic parameters of the individuals accurately match the target electromagnetic parameters, and the individuals are efficiently and finely optimized and adjusted, so that the individuals in the population quickly converge to form a target individual and obtain a target electromagnetic structure. Through the iterative processing process, the population is quickly and efficiently processed, the data complexity is reduced, and the optimization efficiency and speed are improved.

[0016] Therefore, by the pixelated electromagnetic structure design optimization method of the embodiment, the pixelated optimization design process realizes the multi-layer target electromagnetic structure. In this way, the high-dimensional and high-degree design space brought by pixelation can be utilized, and the initial start-up stage of the optimization of the electromagnetic structure is ensured to generate a basic conductive path with physical meaning, realize accurate control of complex electromagnetic performance, and guarantee the performance upper limit of the electromagnetic structure. At the same time, the design optimization efficiency, reliability and practicability of the electromagnetic structure are greatly improved, and the universality and generality of the design optimization of the electromagnetic structure are improved.

[0017] Next, the specific implementation steps of the pixelated electromagnetic structure design optimization method provided by the embodiment will be described in detail in combination with Figure 1 and Figure 2 . First, step S101 is performed to obtain the design parameters of the target electromagnetic structure, including the process parameters of the metal layers in the target electromagnetic structure, the pixel grid map of the metal layers, and the target electromagnetic parameters of the target electromagnetic structure.

[0018] Specifically, the running device of the entire design optimization method of the embodiment is a personal computer, a tablet computer or other electronic devices. The running environment on the running device establishes a communication connection with the electromagnetic simulation software (such as HFSS, XMC, Momentum) through an automatic script interface (such as the pywin32 library of Python). The running device and the running environment can be set according to actual needs.

[0019] The target electromagnetic structure can be a single metal layer electromagnetic structure or a multi-layer electromagnetic structure. The process parameters of the metal layers in the target electromagnetic structure include the physical size of the target electromagnetic structure, i.e. the physical size of each metal layer of the target electromagnetic structure, the thickness of each metal layer and the conductivity of each metal layer, and the dielectric constant and thickness of the interlayer dielectric between the metal layers. The pixel grid map of the metal layer is based on the physical size of the metal layer to form a QxQ two-dimensional pixel grid map by discretizing the metal layer, and each metal layer corresponds to a pixel grid map. Q can be set to a specific value according to actual needs. On the pixel grid map of the metal layer, a pixel grid represents a pixel block of the metal layer corresponding to the pixel grid. The target electromagnetic parameters of the target electromagnetic structure can be set according to actual needs, and the target electromagnetic parameters are usually the impedance conversion ratio of the transformer. Alternatively, the target electromagnetic parameters are the impedance of a specific port, i.e. the target input impedance.

[0020] The design parameters further include a target tolerance ε, a population size N (i.e. the number of individuals in the population), a maximum number of iteration processes G_max, and a crossover probability Pc and a mutation probability Pm.

[0021] Each metal layer's pixel grid is established based on a Cartesian coordinate system. Each metal layer is placed parallel to the XOY plane, and multiple metal layers are overlapped. All pixels within the same metal layer have the same Z-coordinate. When describing the coordinate positions of pixel blocks later, we will mainly focus on the horizontal and vertical coordinates; different metal layers represent different vertical coordinates.

[0022] It should be noted that, in order to illustrate the entire optimization design method, the electromagnetic structure of two metal layers—one high-metal layer and one low-metal layer—will be used as an example. In this example, such as... Figure 3 As shown, the process parameters of the target electromagnetic structure include: the physical dimensions of each metal layer (high metal layer and low metal layer), the thickness and conductivity of each metal layer, and the dielectric constant and thickness of the interlayer dielectric between the metal layers. The pixel grid diagram of the metal layers is obtained by discretizing the metal layers into a Q×Q two-dimensional pixel grid diagram based on their physical dimensions, thus yielding the pixel grid diagrams of the high metal layer and the low metal layer. On the pixel grid diagram of the metal layers, a pixel grid represents a pixel block of the corresponding metal layer. The target electromagnetic parameter of the target electromagnetic structure is the target impedance. Alternatively, the target electromagnetic parameter is the impedance of a specific port. For example, a transformer has two ports. If port 1 is connected to a load, the impedance of port 2 needs to be optimized; in this case, port 2 is a specific port. Conversely, if port 2 is connected to a load, the impedance of port 1 needs to be optimized; in this case, port 1 is a specific port. In the pixel grid diagram of the low metal layer, the coordinates of each pixel block are (X1, Y1, Z1), (X1, Y2, Z1), ..., (XQ, YQ, Z1). In the pixel grid diagram of the high metal layer, the coordinates of each pixel block are (X1,Y1,Z2), (X1,Y2,Z2), ..., (XQ,YQ,Z2).

[0023] In this embodiment, the design parameters of the target electromagnetic structure mainly include the process parameters of the metal layer of the target electromagnetic structure, the pixel mesh map of the metal layer, and the target electromagnetic parameters of the target electromagnetic structure. This demonstrates that the entire design optimization method of this embodiment does not limit the specific structure and form of the target electromagnetic structure, increasing the diversity of the target electromagnetic structure, improving its versatility and universality, and enhancing the generalizability of the design optimization process.

[0024] Next, step S102 is executed to generate an initial population based on the target electromagnetic parameters. Each individual in the initial population carries a list of key points, and the key points in the list of key points of each individual are the key pixels on the pixel grid map of each individual.

[0025] Specifically, based on the target electromagnetic parameters, N keypoint lists are obtained. Here, N is an integer greater than 1. Each keypoint list represents the total set of keypoints for an individual, and includes multiple sublists. Each sublist represents the set of keypoints for the topological structure of a pixel mesh map within a single metal layer of that individual. Taking an electromagnetic structure (i.e., an individual) consisting of two metal layers (a high-metal layer and a low-metal layer) as an example, one keypoint list represents the keypoint list for an individual of that electromagnetic structure, i.e., the total set of keypoints for that individual. N keypoint lists represent the keypoint lists for N individuals of that electromagnetic structure, i.e., the total set of keypoints for the N individuals of that electromagnetic structure. An individual's keypoint list includes two sublists: one sublist represents the set of keypoints for the topological structure of the pixel mesh map within the high-metal layer, and the other sublist represents the set of keypoints for the topological structure of the pixel mesh map within the low-metal layer. Within each sublist, the keypoints are ordered sequentially from the keypoint corresponding to the starting point of the port to the keypoint corresponding to the ending point of the port. This sequential order can be counter-clockwise, clockwise, or customized according to specific requirements.

[0026] The specific process of obtaining a list of N key points based on the target electromagnetic parameters is as follows: Based on the target electromagnetic parameter Z target and target electromagnetic parameters Z target Corresponding load impedance Z load Determine the impedance transformation ratio Ratio = Re(Z) target ) / Re(Z load For example, the target electromagnetic parameter is the input impedance Z of port 1 of the transformer. target If the load is connected to port 2 of the transformer, then the impedance of port 2 is the load impedance Z corresponding to the target electromagnetic parameters. load After obtaining the impedance transformation ratio, the physical lengths of the primary and secondary coils are determined based on this ratio and the turns ratio between the primary and secondary coils. For example, in a resonant transformer architecture where both primary and secondary coils are single-turn (1:1), the turns ratio is 1:1. Using electromagnetic approximations (the inductance of a single-turn coil is approximately proportional to its physical length, and the impedance ratio is approximately equal to the inductance ratio), the empirical formula L1 / L2 ≈ Ratio can be derived, where L1 is the physical length of the primary coil and L2 is the physical length of the secondary coil. Based on this empirical formula, the physical lengths L1 and L2 of the primary and secondary coils are calculated.

[0027] Based on the physical lengths of the primary and secondary coils, N keypoint lists with different topologies and varying numbers of keypoints are randomly generated. Different topologies include square, circular, and octagonal topologies, etc. The length of each keypoint list is unlimited, meaning the number of keypoints in each list is unlimited. For example, ... Figure 4 As shown, a square topology with a side length of 12 pixels is randomly generated for this individual. Both the high-metal and low-metal layers of this individual have square topologies. The coordinates of the keypoints in the high-metal layer's sublist are P1(x1, y1), P2(x2, y2), P3(x3, y3), P4(x4, y4), P5(x5, y5), P6(x6, y6). The coordinates of the keypoints in the low-metal layer's sublist are P1'(x1', y1'), P2'(x2', y2'), P3'(x3', y3'), P4'(x4', y4'), P5'(x5, y5'), P6'(x6, y6'). The keypoints in the high-metal layer's sublist are ordered from the keypoint corresponding to the port starting point (i.e., as shown in the image). Figure 4 The key points of the high-metal layer shown are arranged in a clockwise order from point P1 to the key point corresponding to the end of the port (i.e., as shown in the figure). Figure 4 The key points of the high-metal layer (P6) are arranged in the following order. The key points of the sublist of low-metal layers are arranged in the same order, and will not be repeated here.

[0028] In this embodiment, N distinct key point lists are randomly generated based on the target electromagnetic parameters, i.e., N distinct individuals. This pixelated metal layer provides a high-dimensional, high-degree-of-freedom design space, increasing the diversity of individuals and enabling the exploration of superior individuals, i.e., the target electromagnetic structure that satisfies the target electromagnetic parameters.

[0029] After obtaining N keypoint lists, for each keypoint list, in each pixel grid of an individual, firstly, based on the sublist corresponding to the pixel grid, perform path search processing on the keypoint set corresponding to the sublist to obtain the conductive skeleton path between keypoints in the keypoint set. Then, based on the conductive skeleton path, obtain the fine-tuning region of the conductive skeleton path. Then, based on the keypoint set, the conductive skeleton path, and the fine-tuning region, obtain the pixel structure of the pixel grid. Perform the above path search processing and fine-tuning region operation on each pixel grid of an individual to obtain the pixel structure of each pixel grid of the individual, thus obtaining the individual.

[0030] Perform the pixel mesh graph processing described above on each keypoint list to obtain N individuals, thus forming the initial population. The pixel mesh graph processing involves creating an individual based on its individual pixel mesh graph.

[0031] In each pixel grid map of each metal layer in the individual corresponding to each key point list, a path search process is performed on the key point set corresponding to the sub-list to obtain a specific process of the conductive skeleton path between the key points in the key point set: The first key point in the key point set is taken as the starting point of the current path segment, and the second key point in the key point set is taken as the end point of the current path segment. The A method is used to search for the current path segment. Then, the second key point in the key point set is taken as the starting point of the next path segment of the current path segment, and the third key point in the key point set is taken as the end point of the next path segment of the current path segment. The A method is used to search for the next path segment of the current path segment, until the last key point in the key point set is taken as the end point. Each path segment is searched segment by segment to obtain the conductive skeleton path. Each path segment is A method.

[0032] A The total cost function of the A method is shown in formula (1): f( ) = g( ) + h( ) (1).

[0033] f( ) is the total cost function value of the A method, i.e., the total cost function value of the current path segment, g( ) is the actual cost of the current path segment, h( ) is the estimated cost from the end point of the current path segment to the end point of the next path segment (i.e., the next path segment of the current path segment), is the end point of the current path segment, is the end point of the next path segment of the current path segment, is the horizontal coordinate of the end point of the current path segment, is the horizontal coordinate of the end point of the next path segment of the current path segment, is the vertical coordinate of the end point of the current path segment, is the vertical coordinate of the end point of the next path segment of the current path segment. The estimated cost is represented by the Manhattan distance, so that the path segment is a path segment with only right-angle corners.

[0034] It should be noted that the A method searches for the path segment, and when the A method searches to the boundary of the pixel grid map of the metal layer, it is determined that an obstacle is encountered, and the obstacle is bypassed to continue searching. Therefore, the A The method does not search for a path outside the pixel grid map boundary of the metal layer. Also, A The method tends to go straight in the absence of obstacles, thereby efficiently finding the shortest path between two key points, ensuring the stability and reliability of each path segment, and improving the design optimization efficiency of individuals.

[0035] In the pixel grid map of each metal layer in the individual corresponding to each key point list, based on the conductive skeleton path, the specific process of obtaining the fine-tuning region of the conductive skeleton path is: Along each pixel grid on the conductive skeleton path, a preset number of pixel grids are expanded according to a preset range to obtain the fine-tuning region. The preset range and the preset number can be set according to actual needs.

[0036] Specifically, taking the key point list of the individual as shown in Figure 4 After obtaining the key point list, the key point set of the high metal layer (i.e., the sub-list of the high metal layer) is subjected to path search processing in the pixel grid map of the high metal layer of the individual, to obtain the conductive skeleton path between the key points of the high metal layer. The sub-list of the high metal layer is {P1(x1, y1), P2(x2, y2), P3(x3, y3), P4(x4, y4), P5(x5, y5), P6(x6, y6)}.

[0037] The conductive skeleton path between the key points of the high metal layer is obtained as follows: As shown in Figure 4 , the key point P1 (i.e., the port starting point) is taken as the starting point of the current path segment, and the key point P2 is taken as the end point of the current path segment. The path segment P1-2 is determined by A The total cost function of the method. The current path segment P1-2 is A The path segment with the minimum total cost function value of the method. The next path segment of the current path segment P1-2 is taken as the path segment P2-3, and the starting point of the path segment P2-3 is the key point P2, and the end point of the path segment P2-3 is the key point P3. The path segment P2-3 is determined by A The total cost function of the method. The path segment P2-3 is A The path segment with the minimum total cost function value of the method. The next path segment is P3-4, and the starting point of the path segment P3-4 is the key point P3, and the end point of the path segment P3-4 is the key point P4. The path segment P3-4 is determined by A The total cost function of the method. The path segment P3-4 is A The path segment with the minimum total cost function value of the method. In this way, each path segment is searched step by step until the last key point P6 in the sub-list of the high metal layer is reached as the end point, and each path segment is respectively path segment P1-2, path segment P2-3, path segment P3-4, path segment P4-5, and path segment P5-6. Each path segment is combined to obtain the conductive skeleton path of the high metal layer.

[0038] Based on the conductive skeleton path between the key points of the high metal layer, a fine-tuning region of the conductive skeleton path of the high metal layer is obtained. The specific obtaining process of the fine-tuning region of the high metal layer is as follows: Figure 4 As shown in the conductive skeleton path of the high metal layer, along each pixel grid on the conductive skeleton path of the high metal layer, a preset number of 3 pixel grids are expanded in the preset range direction on the left and right sides to obtain a fine-tuning region, that is, Figure 4 a gray pixel grid as shown.

[0039] The key points of the sub-list of the high metal layer (i.e., the key point set of the high metal layer), the conductive skeleton path of the high metal layer, and the fine-tuning region of the high metal layer form the pixel architecture of the pixel grid map of the high metal layer.

[0040] Similarly, in the pixel grid map of the low metal layer of the individual, the key point set of the low metal layer (i.e., the sub-list of the low metal layer) is subjected to path search processing to obtain the conductive skeleton path between the key points of the low metal layer, that is, the conductive skeleton path of the low metal layer. Based on the conductive skeleton path of the low metal layer, a fine-tuning region of the conductive skeleton path of the low metal layer is obtained, that is, the fine-tuning region of the low metal layer. The key points of the sub-list of the low metal layer (i.e., the key point set of the low metal layer), the conductive skeleton path of the low metal layer, and the fine-tuning region of the low metal layer form the pixel architecture of the pixel grid map of the low metal layer. The obtaining process of the conductive skeleton path of the low metal layer, the obtaining process of the fine-tuning region, and the obtaining process of the pixel architecture are the same as those of the conductive skeleton path of the high metal layer, the obtaining process of the fine-tuning region, and the obtaining process of the pixel architecture, and will not be described again.

[0041] After obtaining the pixel architecture of the high metal layer and the pixel architecture of the low metal layer, the individual is formed based on the pixel architecture of the high metal layer and the pixel architecture of the low metal layer.

[0042] Each individual in the initial population performs the pixel grid map processing described above (i.e., the obtaining process of the individual as shown in Figure 4 ), to obtain N individuals, and then to obtain the initial population.

[0043] In step S102, the key point list of each individual obtained according to the target electromagnetic parameter is used as the optimization gene of each individual, that is, the key points in the key point list of each individual form the macroscopic skeleton of the individual. Then, based on the key point list of each individual, that is, the key points of the pixel grid map of each metal layer of each individual, the A The method searches for the conductive skeleton path of each metal layer. The A The method constructs the conductive skeleton path to ensure the physical connectivity of the key points, so that the structure of the individual has a continuous conductive path, and thus the individual forms an effective electromagnetic coupling structure. Then, the fine-tuning region is obtained based on the conductive skeleton path of each metal layer. Through the setting of the fine-tuning region, an effective basis is provided for subsequent optimization of the individual, and the individual is finely optimized to accurately match the target electromagnetic parameter, thereby improving the design optimization efficiency of the individual. According to the key point list, the conductive skeleton path and the fine-tuning region of each metal layer of each individual, the specific structure of each individual is obtained to make each individual form an effective electromagnetic coupling structure, thereby further improving the design optimization efficiency of each individual. Therefore, a plurality of individuals are randomly generated according to the target electromagnetic parameter of the target electromagnetic structure, and then an initial population is obtained. These individuals are completely different, providing a rich sample of individuals that meet the target electromagnetic parameter for finding the target electromagnetic structure, improving the diversity and automation of the design optimization process of the target electromagnetic structure, reducing the limitations of the design optimization process, and improving the design optimization efficiency and practicality.

[0044] Then, step S103 is performed to perform performance evaluation processing on the initial population to obtain the error value of the initial population.

[0045] Specifically, each individual in the initial population is encoded to obtain a pixel encoding matrix of each individual. The specific process of encoding is as follows: the pixel grid map of each metal layer of each individual in the population is compiled into a pixel encoding matrix of preset encoding values to obtain a pixel encoding matrix of each metal layer of each individual, and then a pixel encoding matrix of each individual is obtained. The dimensions of the pixel grid map are consistent with the dimensions of the corresponding pixel encoding matrix. The pixel grid in the pixel grid map corresponds to the pixel point in the pixel encoding matrix one by one. The form and specific value of the preset encoding value can be set according to actual needs. The encoding value of the pixel grid corresponding to the key point of the pixel grid map, the encoding value of the pixel grid corresponding to the conductive skeleton path of the pixel grid map, and the encoding value of the pixel grid corresponding to the fine-tuning region of the pixel grid map, as well as the encoding value of the pixel grid corresponding to the region other than the key point, the conductive skeleton path and the fine-tuning region of the pixel grid map can be set according to actual needs.

[0046] An example is that the pixel grid map of each metal layer of each individual in the population is compiled into a pixel encoding matrix in which the pixel points are represented by preset encoding values of 0 and 1. For example, the pixel grid map of each metal layer of each individual in the population is compiled into a pixel encoding matrix in which the pixel points are represented by preset encoding values of 0 and 1. Figure 5shown Figure 4 The pixel mesh diagram shown is a high-metal layer pixel map. This pixel mesh diagram includes a keypoint set (i.e., a sublist of high-metal layers), conductive skeleton paths, and fine-tuning areas for the high-metal layer, as well as other regions besides these. During the compilation of this pixel mesh diagram, the pixel meshes corresponding to the high-metal layer keypoint set (i.e., a sublist of high-metal layers), conductive skeleton paths, and fine-tuning areas are compiled as pixels with a value of 1, while the pixel meshes corresponding to other regions of the high-metal layer are compiled as pixels with a value of 0. Pixels with a value of 1 represent pixels in a metallic state, and pixels with a value of 0 represent pixels in a non-metallic state. The pixel mesh diagram of the high-metal layer has a dimension of Q×Q, and the compiled pixel encoding matrix of the high-metal layer also has a dimension of Q×Q, i.e., a pixel encoding matrix consisting of Q×Q pixels. Therefore, only 0 and 1 are used to represent the preset encoded values ​​in the Q×Q pixel encoding matrix of the high-metal layer. Similarly, the encoding process of the pixel mesh diagram of the low-metal layer is the same as that of the high-metal layer pixel mesh diagram, and will not be described again. The pixel encoding matrix of an individual is obtained based on the pixel encoding matrix of the high metal layer and the pixel encoding matrix of the low metal layer.

[0047] By encoding each individual in the population, a pixel encoding matrix is ​​obtained for each individual. This facilitates electromagnetic simulation of each individual, enabling the efficient and rapid creation of the corresponding 3D metallic entity during the electromagnetic simulation process, and allowing for the efficient and accurate execution of full-wave electromagnetic simulation operations. This ensures the stability, reliability, and realism of each individual, improving the efficiency of electromagnetic structure design and optimization.

[0048] After obtaining the pixel encoding matrix of each individual, the actual electromagnetic parameters of each individual are obtained based on the pixel encoding matrix. Specifically, the actual electromagnetic parameters correspond to the target electromagnetic parameters. Using electromagnetic simulation software, a three-dimensional metal entity and simulation parameters of each individual are established based on the pixel encoding matrix. Based on the three-dimensional metal entity and simulation parameters of each individual, the actual electromagnetic parameters of each individual are simulated and obtained. That is, the target electromagnetic parameters are the target input impedance, and the actual electromagnetic parameters are the actual input impedance.

[0049] Based on the actual electromagnetic parameters and target electromagnetic parameters of each individual, the error value of each individual is obtained, and then the error value of the initial population is obtained. Based on the error value of each individual, the fitness of each individual is obtained.

[0050] Specifically, for each individual, based on the individual's actual electromagnetic parameters Z sim and target electromagnetic parameters Z target The individual error value Error is obtained, as shown in formula (3).

[0051] (3).

[0052] In the error value function, Re represents a real part value of the electromagnetic parameter of the individual, i.e. is a real part value of the actual electromagnetic parameter of the individual, is a real part value of the target electromagnetic parameter, and Im is an imaginary part value of the electromagnetic parameter of the individual, i.e. is an imaginary part value of the actual electromagnetic parameter of the individual, is an imaginary part value of the target electromagnetic parameter. and are a real part weight and an imaginary part weight respectively set according to actual requirements.

[0053] Each individual obtains an error value of each individual based on formula (3), and then forms the error values of the initial population.

[0054] In addition, for each individual, the fitness of the individual is obtained according to the error value of the individual, and is specifically shown in formula (4): (4).

[0055] wherein, is the fitness of the individual, is a constant set according to actual requirements.

[0056] Each individual obtains the fitness of each individual based on formula (4).

[0057] Whether there is a target individual, i.e. a target electromagnetic structure, in the population is judged by the error value of each individual in the population. In this way, each individual is evaluated, the quality of each individual is efficiently quantified, and whether there is a target electromagnetic structure in the population is quickly and simply and efficiently judged, and the design optimization efficiency is further improved.

[0058] After obtaining the error value of each individual of the initial population, if the error value of each individual in the initial population is not located in the target error range, it is determined that the error value of the initial population is not located in the target error range, which indicates that there is no target individual in the initial population. The target error range can be set according to actual requirements. If the error value of a certain individual in the initial population is located in the target error range, it is determined that the error value of the initial population is located in the target error range, which indicates that there is a target individual in the initial population.

[0059] If the target individual does not exist in the initial population, step S104 is performed. If the error value of the initial population is not within the target error range, micro-optimization processing and macro-optimization processing are performed on each individual in the initial population based on the key point list of the individual to obtain a current population. The current population is then subjected to performance evaluation processing to implement iterative processing of the current population until the error value of the current population is within the target error range or the number of iterations of the current population reaches a preset number. The current population is determined as the target population.

[0060] Specifically, if the target individual does not exist in the initial population, micro-optimization processing and macro-optimization processing are performed on the initial population to obtain a current population. The specific process of the micro-optimization processing of the initial population is as follows. Step (1), the individuals in the initial population are sorted in descending order of fitness, and M1 individuals are selected from the initial population starting from the individual with the highest fitness, where M1 is an integer greater than 1. The specific value of M1 is set according to actual needs. For example, assume that there are 10 individuals in the initial population, which are individual A1, individual A2, individual A3,..., and individual A10. The 10 individuals are sorted in descending order of fitness. The fitness of the 10 individuals is ranked as follows: 9 / 10 (individual A8), 8 / 10 (individual A3), 7 / 10 (individual A2), 6 / 10 (individual A10), 5 / 10 (individual A1), 5 / 10 (individual A4), 4 / 10 (individual A5), 3 / 10 (individual A7), 2 / 10 (individual A6), and 2 / 10 (individual A9). M1 is set to 2, and two individuals are selected from the 10 individuals. Starting from the individual with the highest fitness, the two individuals are individual A8 and individual A3.

[0061] Step (2), after selecting M1 individuals, for each individual in the M1 individuals, the actual electromagnetic parameters of the individual are compared with the target electromagnetic parameters to obtain a comparison result. Specifically, for each individual in the M1 individuals, the real and imaginary parts of the actual electromagnetic parameters of the individual are compared with the real and imaginary parts of the target electromagnetic parameters to obtain a comparison result of the individual.

[0062] Step (3), according to the comparison result, a candidate pixel point of the pixel encoding matrix of the individual is determined, and the candidate pixel point is subjected to state flipping processing to obtain a micro-optimized individual.

[0063] Specifically, the comparison result one is: if the imaginary part difference between the imaginary part value of the actual electromagnetic parameter of the individual and the imaginary part value of the target electromagnetic parameter is not located in the imaginary part difference threshold range, and the imaginary part value of the actual electromagnetic parameter of the individual is greater than the imaginary part value of the target electromagnetic parameter, a first pixel point in the pixel coding matrix of the individual is searched out, which satisfies the electric field intensity condition and is in the non-metal state, the first pixel point is taken as the candidate pixel point of the individual, and metal is added on the candidate pixel point of the individual, so that the candidate pixel point of the individual is in the metal state, until the imaginary part difference is located in the imaginary part difference threshold range, and the individual after micro-optimization is obtained. The imaginary part difference threshold range is set according to actual requirements.

[0064] In the case that the imaginary part difference between the imaginary part value Im(Z sim ) of the actual electromagnetic parameter of the individual and the imaginary part value Im(Z target ) of the target electromagnetic parameter is not located in the imaginary part difference threshold range, and Im(Z sim ) > Im(Z target ), it is indicated that the imaginary part value Im(Z sim ) of the actual electromagnetic parameter of the individual is too large, the electromagnetic structure corresponding to the individual is inductive, and the equivalent capacitance needs to be increased. In this case, any one pixel coding matrix of the individual is traversed, and a current electric field pixel point is searched out in the pixel coding matrix. The electric field pixel point is the pixel point with the maximum electric field intensity in each search process, that is, the current electric field pixel point is the pixel point with the maximum electric field intensity in the current search process. It should be noted that, in the traversal of the pixel coding matrix of the individual, the traversal range is the pixel point range of the pixel coding matrix except the key point and the conductive skeleton path, so as to avoid the existence of the open circuit structure of the electromagnetic structure of the individual. The traversal order is to traverse the fine tuning area of the pixel coding matrix first, and then gradually expand the traversal range along the fine tuning area.

[0065] If the current electric field pixel point is the pixel point of the fine tuning area of the pixel coding matrix, and the specified electric field pixel point corresponding to the current electric field pixel point is in the non-metal state, the specified electric field pixel point is taken as the first pixel point, metal is added on the first pixel point, so that the specified electric field pixel point is in the metal state, and then the imaginary part difference is continuously obtained. The specified electric field pixel point is the pixel point at the position corresponding to the position of the current electric field pixel point in the pixel coding matrix except the pixel coding matrix. The current electric field pixel point is the pixel point of the fine tuning area of the pixel coding matrix, which indicates that the current electric field pixel point is the pixel point in the metal state. After the imaginary part difference is continuously obtained, it is judged whether the imaginary part difference is located in the imaginary part difference threshold range. If the imaginary part difference is not located in the imaginary part difference threshold range, and Im(Z sim ) > Im(Z target ), the next electric field pixel point is searched out again in the pixel coding matrix, and the next electric field pixel point is taken as the current electric field pixel point for judgment.

[0066] If the current electric field pixel is a pixel in the fine-tuning region, and the corresponding specified electric field pixel is in a metallic state, then the next electric field pixel is searched again in the pixel encoding matrix, and this next electric field pixel is used as the current electric field pixel for judgment. The next electric field pixel is the pixel with the largest electric field strength after excluding the current electric field pixel during the second search process.

[0067] In this case, if the current electric field pixel is a pixel in a non-fine-tuning region, indicating that the current electric field pixel is in a non-metallic state, and the corresponding specified electric field pixel is also in a non-metallic state, then both the current electric field pixel and the specified electric field pixel are taken as the first pixel, and metal is added to the first pixel so that both the current electric field pixel and the specified electric field pixel are in a metallic state, and then the imaginary part difference is obtained again. If the current electric field pixel is a pixel in a non-fine-tuning region, and the corresponding specified electric field pixel is in a metallic state, then the current electric field pixel is taken as the first pixel, and metal is added to the first pixel so that the current electric field pixel is in a metallic state, and then the imaginary part difference is obtained again. It should also be noted that the electric field strength condition is the condition for finding the pixel with the largest electric field strength in each search process of the pixel encoding matrix.

[0068] The imaginary part of the actual electromagnetic parameters of an individual, Im(Z). sim The imaginary part of the target electromagnetic parameters, Im(Z), is related to the target electromagnetic parameters. target The imaginary part difference between ) is not within the imaginary part difference threshold range, and Im(Z) sim > Im (Z) target In the case of ), during the process of traversing the individual's pixel encoding matrix, the current electric field pixel is gradually searched out and the first pixel is obtained. Metal is added to the first pixel until the imaginary part difference is within the imaginary part difference threshold range, and the micro-optimized individual is obtained, that is, the individual forms the individual with added capacitance.

[0069] For example, such as Figure 6a and Figure 6b shown Figure 4 The local image near keypoint P4 of the pixel coding matrix of the high-metal layer, showing the imaginary part of the actual electromagnetic parameters Im (Z) of the individual. sim The imaginary part of the target electromagnetic parameters, Im(Z), is related to the target electromagnetic parameters. target The imaginary part difference between ) is not within the imaginary part difference threshold range, and Im(Z) sim > Im (Z) targetIn the case of ), traverse the pixel encoding matrix of the individual's high metal layer, and search for the pixel with the largest electric field intensity B1 in the individual's high metal layer pixel encoding matrix, that is, B1 is the current electric field pixel. At this time, if Figure 6a As shown, B1 is a pixel in the fine-tuning region of the pixel encoding matrix. A pixel with a value of 1 in B1 indicates a pixel in a metallic state. It is necessary to determine whether the pixel corresponding to the specified electric field, B1', is a pixel in a non-metallic state. B1' is the pixel in the pixel encoding matrix corresponding to the position of B1 in the low-metal layer. That is, B1 and B1' have the same coordinate positions in the pixel encoding matrix, limited to the horizontal and vertical coordinates. For example, B1's coordinate position in the pixel encoding matrix of the high-metal layer is (3,3), meaning it is located at the 3rd row and 3rd column. Similarly, B1''s coordinate position in the pixel encoding matrix of the low-metal layer is also (3,3), meaning it is located at the 3rd row and 3rd column.

[0070] If B1' is in a non-metallic state, then B1' is taken as the first pixel, and metal is added to B1' to make it metallic, and the imaginary part difference is then obtained. If B1' is metallic, then the next electric field pixel B2 is searched again in the pixel encoding matrix of the high-metal layer, and B2 is used as the current electric field pixel for judgment. In the process of searching for B2, after excluding B1, the pixel with the largest electric field intensity is B2. After obtaining B2, if B2 is a pixel in the fine-tuning region of the pixel encoding matrix, it means that B2 is a pixel with a value of 1, and then B2 is optimized according to the optimization process of B1, which will not be elaborated further. Figure 6b As shown, if B2 is a pixel in the non-fine-tuned region of the pixel encoding matrix of the high metal layer, it means that the pixel with B2 is 0, that is, B2 is a pixel in a metal-free state. It is also necessary to determine whether the pixel B2' corresponding to the specified electric field of B2 is a pixel in a metal-free state. The determination and optimization process of B2' is the same as that of B1', and will not be described again.

[0071] When B1 is searched out, B1 is a pixel point of the non-fine-tuning area of the pixel coding matrix, B1 is a pixel point of 0, i.e. B1 is a pixel point in a non-metallic state. It is needed to judge whether the designated electric field pixel point B1' corresponding to B1 is a pixel point in a non-metallic state. When B1 and B1' are both pixel points in a non-metallic state, B1 and B1' are taken as the first pixel points, and metal is added on B1 and B1' so that B1 and B1' are in a metallic state, and then the imaginary part difference value is continuously obtained. When B1 is in a non-metallic state and B1' is in a metallic state, only B1 is taken as the first pixel point, and metal is added on B1 so that B1 is in a metallic state, and then the imaginary part difference value is continuously obtained. Through the above optimization process, the imaginary part difference value is located in the imaginary part difference value threshold range, so that Figure 4 The individual shown becomes the micro-optimized individual.

[0072] The comparison result two: if the comparison result is that the imaginary part difference value is not located in the imaginary part difference value threshold range, and the imaginary part value of the actual electromagnetic parameter of the individual is less than the imaginary part value of the target electromagnetic parameter, a second pixel point in a metallic state and satisfying the electric field intensity condition is searched out in the pixel coding matrix of the individual, and then the second pixel point is taken as the candidate pixel point of the individual, and the metal is removed from the candidate pixel point of the individual so that the candidate pixel point of the individual is in a non-metallic state, until the imaginary part difference value is located in the imaginary part difference value threshold range, and the micro-optimized individual is obtained.

[0073] In the case that the imaginary part difference value between the imaginary part value Im(Z sim ) of the actual electromagnetic parameter of the individual and the imaginary part value Im(Z target ) of the target electromagnetic parameter is not located in the imaginary part difference value threshold range, and Im(Z sim ) < Im(Z target ), it is indicated that the imaginary part value Im(Z sim ) of the actual electromagnetic parameter of the individual is too small, and the electromagnetic structure corresponding to the individual is capacitive, and the equivalent capacitance needs to be reduced. In this case, any one pixel coding matrix of the individual is traversed, and a current electric field pixel point is searched out in the pixel coding matrix. The current electric field pixel point is taken as the second pixel point, the metal is removed from the second pixel point (i.e. the current electric field pixel point) so that the second pixel point (i.e. the current electric field pixel point) is in a non-metallic state, and the imaginary part difference value is continuously obtained. The setting requirement of the current electric field pixel point in the case of the comparison result two is consistent with the setting requirement of the current electric field pixel point in the case of the comparison result one, i.e. the current electric field pixel point is the pixel point with the maximum electric field intensity in each search process. The traversal range and the traversal order in the case of the comparison result two are also consistent with the traversal range and the traversal order in the case of the comparison result one, and are not described herein.

[0074] After the imaginary part difference is obtained, it is determined whether the imaginary part difference is located in the imaginary part difference threshold range. If the imaginary part difference is not located in the imaginary part difference threshold range, and Im(Z sim ) < Im(Z target ), the next electric field pixel point is searched again in the pixel coding matrix, and the next electric field pixel point is taken as the current electric field pixel point. The next electric field pixel point is the pixel point with the maximum electric field intensity in the re-searching process, excluding the current electric field pixel point. The next electric field pixel point in the second comparison result is consistent with the next electric field pixel point in the first comparison result. After the next electric field pixel point is obtained, the current electric field pixel point is taken as the second pixel point, the metal is removed from the second pixel point (i.e. the current electric field pixel point) to make the second pixel point (i.e. the current electric field pixel point) in a metal-free state, and the imaginary part difference is continuously obtained. Until the imaginary part difference is located in the imaginary part difference threshold range, the micro-optimized individual is obtained, that is, the individual forms an individual with reduced capacitance.

[0075] Therefore, in the first comparison result, the metal is added to the same coordinate position on each pixel coding matrix, so that the pixel point at the same coordinate position is in a metal state, the electric field intensity is enhanced, and the purpose of increasing the capacitance is achieved. In the second comparison result, if the metal exists only in a single pixel coding matrix at a certain coordinate position, that is, only the pixel point at the coordinate position is in a metal state. This indicates that the electric field intensity at the coordinate position is weakened, not enhanced, and the purpose of reducing the capacitance is achieved.

[0076] The third comparison result: if the real part difference is not located in the real part difference threshold range, and the real part value of the actual electromagnetic parameter of the individual is greater than the real part value of the target electromagnetic parameter, the third pixel point adjacent to the conductive skeleton path and the fine-tuning region of the individual and satisfying the current density condition is searched in the pixel coding matrix of the individual, the third pixel point is taken as the candidate pixel point of the individual, and the metal is added to the candidate pixel point of the individual to make the candidate pixel point of the individual in a metal state, until the real part difference is located in the real part difference threshold range, and the micro-optimized individual is obtained. The real part difference threshold range is set according to actual requirements.

[0077] The real part value of the actual electromagnetic parameter of the individual is greater than the real part value of the target electromagnetic parameter , that is, the real part difference between the real part value of the actual electromagnetic parameter of the individual and the real part value of the target electromagnetic parameter In this case, traverse any pixel encoding matrix of the individual and search for the current current pixel within that matrix. The current current pixel is the pixel with the highest current density during each search; that is, the current current pixel is the pixel with the highest current density during the current search. It should be noted that the traversal range within the individual's pixel encoding matrix includes all pixels except for keypoints. The traversal order is to first traverse the fine-tuning region and conductive skeleton path of the pixel encoding matrix, and then gradually expand the traversal range along the fine-tuning region.

[0078] If the current current pixel is within the fine-tuning region of the pixel encoding matrix, indicating that the current current pixel is in a metallic state, and the specified current pixel corresponding to the current current pixel is in a non-metallic state, then the specified current pixel is designated as the third pixel, and the specified pixel in the specified region corresponding to the current current pixel is also designated as the third pixel. Metal is added to the third pixel to make it in a metallic state, and the real part difference is continued to be obtained. Here, the specified current pixel corresponding to the current current pixel is the pixel at the position in the pixel encoding matrix other than the current pixel's position. The specified region is the area surrounding the current current pixel's position within the fine-tuning region, and the specified region is adjacent to the current pixel's position within the fine-tuning region. The pixels in the specified region are the first and second specified pixels closest to the current current pixel within the specified region. The first specified pixel is a pixel within the pixel encoding matrix, and the second specified pixel is a pixel located at a position in a pixel encoding matrix other than the first specified pixel. This means that the x-coordinates (X-axis coordinates) and y-coordinates (Y-axis coordinates) of the first and second specified pixels are the same, but their z-coordinates (Z-axis coordinates) are different. Pixels within the specified area are in a non-metallic state.

[0079] If the current current pixel is a pixel within the fine-tuning region of the pixel encoding matrix, and the specified current pixel corresponding to the current current pixel is in a metallic state, it means that both the current current pixel and the specified current pixel are pixels in a metallic state. Then, the specified pixel in the specified region corresponding to the current current pixel is taken as the third pixel, metal is added to the third pixel to make the third pixel in a metallic state, and the real part difference is continued to be obtained.

[0080] After continuing to obtain the real part difference, if the real part value of the actual electromagnetic parameter of the individual... With the real part of the target electromagnetic parameters The real part difference between them is not within the real part difference threshold range, which indicates that there is a deviation in the electromagnetic structure of the individual, and > If the real part difference value between the real part value of the actual electromagnetic parameter of the individual and the real part value of the target electromagnetic parameter is not located in the real part difference value threshold range, and

[0081] the real part value of the actual electromagnetic parameter of the individual and the real part value of the target electromagnetic parameter , the process of traversing the pixel coding matrix of the individual gradually searches for the current current pixel point and obtains a third pixel point, and the metal is added on the third pixel point until the real part difference value is located in the real part difference value threshold range, and a micro-optimized individual is obtained, i.e., the individual becomes a micro-optimized individual. If the real part difference value between the real part value of the actual electromagnetic parameter of the individual and the real part value of the target electromagnetic parameter is not located in the real part difference value threshold range, and

[0082] For example, as shown in a local diagram of a key point P4 in a pixel coding matrix of a high metal layer of Figure 7 , if the real part difference value between the real part value of the actual electromagnetic parameter of the individual and the real part value of the target electromagnetic parameter is not located in the real part difference value threshold range, and Figure 4 In this case, the pixel coding matrix of the high metal layer of the individual is traversed, and the pixel point C1 with the maximum current density is searched in the pixel coding matrix of the high metal layer, i.e., C1 is the current current pixel point. At this time, C1 is a pixel point in the fine-tuning region of the pixel coding matrix of the high metal layer, indicating that the current current pixel point is a pixel point with a metal state, i.e., C1 is a pixel point in a metal state. It is necessary to determine whether the specified current pixel point C1' corresponding to C1 is a pixel point in a non-metal state. C1' is a pixel point at the position of the position of C1 in the pixel coding matrix of the low metal layer, i.e., C1 and C1' have the same horizontal coordinate and vertical coordinate positions in the pixel coding matrix. For example, the coordinate position of C1 in the pixel coding matrix of the high metal layer is (3, 4), i.e., C1 is located at the position of the 3rd row and the 4th column in the pixel coding matrix of the high metal layer. The coordinate position of C1' in the pixel coding matrix of the low metal layer is also (3, 4), i.e., C1' is located at the position of the 3rd row and the 4th column in the pixel coding matrix of the low metal layer.

[0083] ​​​​​If C1' is in the non-metal state, the pixel points C2 and C2' in the designated area corresponding to C1 and C1' are taken as the third pixel points, metal is added to C1', C2 and C2' so that C1', C2 and C2' are in the metal state, and the real part difference value is continuously obtained. If C1' is in the metal state, the pixel points C2 and C2' in the designated area corresponding to C1 are taken as the third pixel points, metal is added to C2 and C2' so that C2 and C2' are in the metal state, and the real part difference value is continuously obtained. The designated area corresponding to the current pixel point is the area adjacent to the fine-tuning area and close to the current pixel point in the fine-tuning area. The pixel point C2 in the designated area is the pixel point closest to the current pixel point in the designated area, and C2 is the pixel point of the pixel coding matrix of the high metal layer. The pixel point C2' in the designated area is the pixel point of the pixel coding matrix of the low metal layer, and the horizontal coordinates and vertical coordinates of C2 and C2' in the pixel coding matrix are consistent. The pixel point C2 in the designated area is 0, that is, C2 is a pixel point in the non-metal state.

[0084] It should be noted that when the real part difference value between the real part value of the actual electromagnetic parameter of the individual and the real part value of the target electromagnetic parameter is not located in the real part difference value threshold range, and > , the current is taken as the conductor of the metal layer, and there is no current in the pixel point in the non-metal state in the metal layer. For example, there is no metal on a pixel point of a certain metal layer of the individual, and the pixel point is not current. Therefore, the current pixel point is not a pixel point in the non-fine-tuning area.

[0085] Comparison result four: if the real part difference value is not located in the real part difference value threshold range, and the real part value of the actual electromagnetic parameter of the individual is greater than the real part value of the target electromagnetic parameter, a fourth pixel point adjacent to the conductive skeleton path and the fine-tuning area of the individual and satisfying the current density condition is searched in the pixel coding matrix of the individual, the fourth pixel point is taken as the candidate pixel point of the individual, and metal is added to the candidate pixel point of the individual to make the candidate pixel point of the individual in the metal state until the real part difference value is located in the real part difference value threshold range, and the micro-optimized individual is obtained.

[0086] When the real part difference value between the real part value of the actual electromagnetic parameter of the individual and the real part value of the target electromagnetic parameter is not located in the real part difference value threshold range, that is, the electromagnetic structure of the individual has deviation, the resistance needs to be adjusted, and < In the case of the fourth comparison result, any one of the pixel coding matrix of the individual is traversed, and the current current pixel point is searched out in the pixel coding matrix. The current current pixel point is taken as the fourth pixel point, and the metal is subtracted from the fourth pixel point (i.e. the current current pixel point) so that the second pixel point (i.e. the current current pixel point) is in a non-metal state, and the real part difference value is continuously obtained. The setting requirements of the current current pixel point in the fourth comparison result case are consistent with the setting requirements of the current current pixel point in the third comparison result case, that is, the current current pixel point is the pixel point with the maximum current density in each search process. The traversal range and traversal order in the fourth comparison result case are also consistent with the traversal range and traversal order in the third comparison result case, and will not be repeated here.

[0087] After the real part difference value is continuously obtained, it is judged whether the real part difference value is located in the real part difference value threshold range. If the real part difference value is not located in the real part difference value threshold range, and < the next current pixel point is searched out again in the pixel coding matrix, and the next current pixel point is taken as the current current pixel point. The next current pixel point is the pixel point with the maximum current density in the search process, excluding the current current pixel point. The setting requirements of the next current pixel point in the fourth comparison result case are consistent with the setting requirements of the next current pixel point in the third comparison result case. After the next current pixel point is obtained, the current current pixel point is taken as the fourth pixel point, the metal is subtracted from the fourth pixel point (i.e. the current current pixel point) so that the fourth pixel point (i.e. the current current pixel point) is in a non-metal state, and the real part difference value is continuously obtained. Until the real part difference value is located in the real part difference value threshold range, the micro-optimized individual is obtained.

[0088] Step (4), each individual in the M1 individuals performs the above comparison processing to the state flip processing of the candidate pixel point of the individual, that is, each individual in the M1 individuals performs the optimization process according to the obtained comparison result, that is, each individual in the M1 individuals performs steps (2)-(4) to obtain M1 micro-optimized individuals.

[0089] In this embodiment, the elite individuals (i.e. M1 individuals) are screened out from the population according to the fitness of the individual. The local pixel points in the fine tuning region of the individual are intelligently guided for fine optimization through the physical field distribution information, i.e. the electric field / current distribution information, of each individual in the M1 individuals, so that the electromagnetic parameters of the elite individuals can accurately match the target electromagnetic parameters. In this way, by retaining the elite individuals in the population and micro-optimizing the elite individuals, the design optimization efficiency, reliability and practicability of the electromagnetic structure are greatly improved, and the generality and universality of the design optimization of the electromagnetic structure are also improved.

[0090] The specific process of the macro-optimization processing of the initial population is: Step 1, M2 individuals in the initial population are selected by roulette selection method, M2 is an integer greater than 1, and N=M1+M2. The specific value of M2 can be set according to actual needs.

[0091] Step 2, based on the crossover probability, M2 individuals are internally crossed to obtain M2 individuals after crossing, and the internal crossing processing is a processing process of exchanging key points in the pixel encoding matrix of the individual that meet the preset cutting point condition.

[0092] Step 3, based on the mutation probability, M2 individuals after crossing are mutated to obtain M2 individuals after mutation, and then M2 macroscopically optimized individuals are obtained, and the mutation processing includes key point adding processing, key point deleting processing or key point moving processing.

[0093] Specifically, in step 1, the specific process of selecting M2 individuals by roulette selection method is as follows: The fitness of each individual in the initial population is added to obtain the total fitness F_total.

[0094] Starting from the first individual in the initial population, the fitness of each individual before the individual is added to the fitness of each individual to obtain the roulette value of each individual. For example, the initial population has 5 individuals, and the fitness of the 5 individuals is f1, f2, f3, f4 and f5 respectively. The roulette value of individual 1 is f1, the roulette value of individual 2 is f1+f2, the roulette value of individual 3 is f1+f2+f3, the roulette value of individual 4 is f1+f2+f3+f4, and the roulette value of individual 5 is f1+f2+f3+f4+f5. The roulette value of the last individual in the initial population is the total fitness.

[0095] According to the interval formed by the total fitness F_total, M2 times of roulette selection processing operation are performed to obtain M2 individuals, wherein each time of roulette selection processing operation includes: A random number R is obtained from the interval [0, F_total], and the selected individual that meets the roulette selection condition is determined according to the random number R and the roulette value of each individual, which is the individual whose roulette value is greater than or equal to the random number R and whose difference between the roulette value and the random number R is the smallest.

[0096] It should be noted that the interval of [0, F_total] is similar to a roulette, and the random number R is similar to a roulette pointer. For example, the roulette value of individual 1 is f1, the roulette value of individual 2 is f1+f2, the roulette value of individual 3 is f1+f2+f3, the roulette value of individual 4 is f1+f2+f3+f4, and the roulette value of individual 5 is f1+f2+f3+f4+f5. The roulette values of individual 3, individual 4, and individual 5 are all greater than or equal to the random number R, but the difference between the roulette value of individual 3 and the random number R is the smallest. Individual 3 is determined as the selected individual. The roulette selection processing operation is performed once each time, and one selected individual is obtained. Then, M2 selected individuals, i.e., M2 individuals, are obtained.

[0097] In this embodiment, the proportion of the fitness value of each individual, i.e., an index for measuring the advantages and disadvantages of the individual, in the sum of the total fitness is taken as the probability of being selected. The higher the fitness of an individual, the larger the interval occupied by the individual on the virtual roulette, and thus the higher the chance of being selected. In this way, the fitness difference of the individual can be effectively amplified, the population evolution can be promoted, and the design optimization efficiency and reliability can be improved.

[0098] In step ②, the probability of each individual in the M2 individuals performing internal crossover processing is the crossover probability Pc. After the internal crossover processing, M2 cross-processed individuals are obtained. The M2 cross-processed individuals include individuals that have undergone internal crossover processing and individuals that have not undergone internal crossover processing. For example, assuming that the crossover probability Pc is 0.8 and the M2 individuals are formed by individual 1, individual 3, and individual 5. The probability of individual 1 undergoing internal crossover processing is 0.8, the probability of individual 3 undergoing internal crossover processing is 0.8, and the probability of individual 5 undergoing internal crossover processing is 0.8.

[0099] The specific process of individual in the M2 individuals performing internal crossover processing is to determine the cut points in the pixel coding matrix (i.e., regarded as a pixel grid graph) of each metal layer of the individual; and exchange the cut points of adjacent metal layers to obtain a cross-processed individual. The cut point is a key point in each pixel coding matrix that meets a preset cut point condition. The preset cut point condition is that the cut point cannot be a key point corresponding to a port, the cut points in the same pixel coding matrix are symmetrically distributed, and the number of cut points is two. The preset cut point condition can also be set according to actual needs. The number of cut points is set to two, so that the individual can reduce the data complexity, more quickly and efficiently optimize, guarantee the reliability and stability in the process of iteration processing, and avoid generating useless electromagnetic structures in the process of iteration.

[0100] For example, as shown in FIG. 6, the pixel coding matrix of the individual is divided into two parts by the cut point, and the two parts are exchanged to obtain the cross-processed individual. Figure 8As shown in the figure, the key points and coordinates of the key points on the high metal layer sub-list (i.e. the pixel grid map or the pixel encoding matrix) of the individual are {P1, P2, P3, P4, P5, P6, P7, P8}. The key points and coordinates of the key points on the low metal layer sub-list (i.e. the pixel grid map or the pixel encoding matrix) of the individual are {P1', P2', P3', P4', P5', P6', P7', P8'}. Assuming that the cut points of the high metal layer are P4 and P5, and the cut points of the low metal layer are P3' and P6'. The cut points of the high metal layer and the cut points of the low metal layer both satisfy the preset cut point condition. The cut points of the high metal layer are P4 and P5, and the cut points of the low metal layer are P3' and P6'. The key points and coordinates of the key points on the new sub-list of the high metal layer of the individual are {P1, P2, P3, P3', P6', P6, P7, P8}, and the key points and coordinates of the key points on the new sub-list of the low metal layer are {P1', P2', P4, P4', P5', P5, P7', P8'}. The vertical axis coordinates of the pixel encoding matrix or the pixel grid map of each metal layer are unchanged. The key points and pixel architecture on the new sub-list of the high metal layer of the individual, and the key points and pixel architecture on the new sub-list of the low metal layer are shown in the figure. Figure 8

[0101] In this embodiment, through internal cross processing, the topological cross fusion of the individual is realized, and the advantages of different designs of the individual are efficiently combined. In this way, the advantages of the individual can be enhanced, such as changing the coupling degree, optimizing the bandwidth or efficiency of the transformer, so that the electromagnetic parameters of the individual efficiently fit the target electromagnetic parameters. In this way, the different advantages of the individual are improved, the diversity of the individual is enhanced, and the universality and generality of the design optimization of the electromagnetic structure are improved, so that more reliable possibilities are provided for the target electromagnetic structure, and the design optimization efficiency and practicality of the electromagnetic structure are improved.

[0102] In step ③, the probability of each of the M2 cross individuals performing mutation processing is the crossover probability Pm. After mutation processing, M2 mutated individuals, i.e. M2 macroscopically optimized individuals, are obtained. The M2 mutated individuals include individuals that have undergone mutation processing and individuals that have not undergone mutation processing. For example, assuming that the crossover probability Pm is 0.9, and the M2 cross individuals are formed by individual 1, individual 3, and individual 5. Among them, individual 3 and individual 5 are individuals that have undergone internal cross processing, and individual 1 is an individual that has not undergone internal cross processing. The probability of individual 1 undergoing mutation processing is 0.9, the probability of individual 3 undergoing mutation processing is 0.9, and the probability of individual 5 undergoing mutation processing is 0.9.

[0103] The specific process of the cross individual in the M2 cross individuals performing mutation processing is as follows: ​The key point addition processing is to add a key point in a fine adjustment region between any two adjacent key points in the pixel coding matrix of the individual after the crossover to obtain a mutated individual.

[0104] As shown in FIG. 4, the key point addition processing is to add a key point in a fine adjustment region between any two adjacent key points in the pixel coding matrix of the individual after the crossover to obtain a mutated individual. Figure 9 As shown in FIG. 4, the key point addition processing is to add a key point in a fine adjustment region between any two adjacent key points in the pixel coding matrix of the individual after the crossover to obtain a mutated individual. Figure 8 As shown in FIG. 4, the key point addition processing is to add a key point in a fine adjustment region between any two adjacent key points in the pixel coding matrix of the individual after the crossover to obtain a mutated individual.

[0105] The key point deletion processing is to randomly delete a key point in the pixel coding matrix of the individual after the crossover, and the deleted key point is a key point not corresponding to a port.

[0106] As shown in FIG. 4, the key point addition processing is to add a key point in a fine adjustment region between any two adjacent key points in the pixel coding matrix of the individual after the crossover to obtain a mutated individual. Figure 10 As shown in FIG. 4, the key point addition processing is to add a key point in a fine adjustment region between any two adjacent key points in the pixel coding matrix of the individual after the crossover to obtain a mutated individual. Figure 8 As shown in FIG. 4, the key point addition processing is to add a key point in a fine adjustment region between any two adjacent key points in the pixel coding matrix of the individual after the crossover to obtain a mutated individual.

[0107] The key point movement processing is to randomly obtain a movement point in the pixel coding matrix of the individual after the crossover, the movement point is a key point to be moved, and the movement point is randomly moved in a preset movement range to obtain a mutated individual. The preset movement range can be set to a range in a fine adjustment region according to actual requirements.

[0108] As shown in FIG. 4, the key point addition processing is to add a key point in a fine adjustment region between any two adjacent key points in the pixel coding matrix of the individual after the crossover to obtain a mutated individual. Figure 11 As shown in FIG. 4, the key point addition processing is to add a key point in a fine adjustment region between any two adjacent key points in the pixel coding matrix of the individual after the crossover to obtain a mutated individual. Figure 8 As shown in FIG. 4, the key point addition processing is to add a key point in a fine adjustment region between any two adjacent key points in the pixel coding matrix of the individual after the crossover to obtain a mutated individual.

[0109] It should be noted that the order of the mutation process or other parameters of the mutation process, such as any one of the processes in the mutation process performed on the crossed individuals (such as only performing the key point addition process or only performing the key point deletion process or only performing the key point movement process) or any combination (such as a combination process of performing the key point addition process and the key point deletion process), and the pixel encoding matrix of a certain metal layer of the crossed individual performs the mutation process or the pixel encoding matrix of the metal layer performs the mutation process or the same position of the pixel encoding matrix of the metal performs the mutation process, can be set according to actual needs.

[0110] In the embodiment, the M2 crossed individuals are randomly disturbed by the mutation probability Pm to obtain M2 mutated individuals, i.e., M2 macroscopically optimized individuals. In this way, diversity is introduced into the individuals in the population, further improving the universality and generality of the design optimization of the electromagnetic structure, providing more reliable possibilities for the target electromagnetic structure, and further improving the design optimization efficiency and practicability of the electromagnetic structure.

[0111] After obtaining the M1 microscopically optimized individuals and the M2 macroscopically optimized individuals, the M1 microscopically optimized individuals and the M2 macroscopically optimized individuals are combined to obtain the current population. The individuals in the current population are N. In the embodiment, in the case where there is no individual in the initial population whose error value meets the target error range, the N individuals in the initial population are subjected to the microscopic optimization process and the macroscopic optimization process, i.e., one iteration is performed to obtain the individuals that are optimized and retain the advantageous pixel architecture. In this way, the actual electromagnetic parameters of the individuals are accurately matched to the target electromagnetic parameters, and the individuals are efficiently and finely optimized and adjusted, so that the individuals in the population quickly converge to form a target individual, and the target electromagnetic structure is obtained. Moreover, by combining macroscopic and microscopic optimization, the electromagnetic structure layout is quickly iterated. Compared with complete pixel randomness, the design data complexity is greatly reduced, and local optimization is prevented. At the same time, the microscopic optimization enables the optimization process to further fit the optimal solution, improves the optimization efficiency and speed, and ensures the quality of the population during iteration.

[0112] After obtaining the current population, steps S103-S104 are performed on the current population. In this way, the iteration process is repeated. Until the error value of the current population is within the target error range or the iteration number of the current population reaches a preset number, the current population is determined as the target population. The preset number of iteration times can be set according to actual needs.

[0113] Then, step S105 is performed to screen a target individual from the target population, and the target individual is determined as the target electromagnetic structure. Specifically, after obtaining the target population, it is found that there are individuals in the target population whose error values are within the target error range. There are multiple individuals in the target population whose error values are within the target error range, and the individual with the minimum error value is taken as the target individual. The electromagnetic structure of the target individual is the target electromagnetic structure.

[0114] Embodiment Two Based on the same inventive concept, the second embodiment of the present application also provides an electromagnetic structure formed by the design optimization method of the pixelated electromagnetic structure described in Embodiment One.

[0115] Electromagnetic structure: This is a general concept, which refers to any physical structure that realizes a specific function (such as energy transmission, signal conversion, impedance transformation, filtering, etc.) through electric and magnetic fields. A simple wire loop, a spiral inductor, a microwave transmission line, can all be called electromagnetic structures.

[0116] In this embodiment, the electromagnetic structure includes a transformer. Transformer: a typical and specialized electromagnetic structure, which is composed of two or more tightly coupled coils (inductors), and works based on the principle of mutual inductance. When the current in one coil (primary coil) changes, it will generate a changing magnetic field, which will pass through another coil (secondary coil), thereby inducing a voltage in it.

[0117] Embodiment Three Based on the same inventive concept, the third embodiment of the present application also provides an electronic device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor implements the steps of any one of the design optimization methods of the pixelated electromagnetic structure when executing the program.

[0118] Those skilled in the art will appreciate that although the preferred embodiments of the present application have been described, those skilled in the art can make further changes and modifications to these embodiments once they know the basic inventive concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications falling within the scope of the present application.

[0119] Obviously, those skilled in the art can make various modifications and variations to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the claims of the present application and their equivalents, the present application is also intended to include these modifications and variations.

Claims

1. A method for designing and optimizing pixelated electromagnetic structures, characterized in that, include: The design parameters of the target electromagnetic structure are obtained, including: the process parameters of the metal layers in the multiple metal layers of the target electromagnetic structure and the pixel grid diagram of the metal layers, and the target electromagnetic parameters of the target electromagnetic structure. Based on the target electromagnetic parameters, an initial population is generated, wherein each individual in the initial population carries a list of key points, and the key points in the list of key points of each individual are key pixels on the pixel grid map of each individual; The initial population is subjected to performance evaluation to obtain the error value of the initial population; If the error value of the initial population is not within the target error range, then based on the key point list of each individual in the initial population, the initial population is subjected to micro-optimization and macro-optimization to obtain the current population. Then, the current population is subjected to performance evaluation to achieve iterative processing of the current population until the error value of the current population is within the target error range or the number of iterations of the current population reaches a preset number, and the current population is determined as the target population. Target individuals are selected from the target population, and the target individuals are identified as the target electromagnetic structures.

2. The design and optimization method for pixelated electromagnetic structures as described in claim 1, characterized in that, The step of generating an initial population based on the target electromagnetic parameters includes: Based on the target electromagnetic parameters, N key point lists are obtained, where N is an integer greater than 1. The key point list is the total set of key points for an individual, and the key point list includes multiple sublists. Each sublist is the set of key points for the topological structure of the pixel mesh map of a metal layer of the individual. For each of the key point lists, in each pixel grid map of the individual, based on the sub-list corresponding to the pixel grid map, a path search process is performed on the key point set corresponding to the sub-list to obtain the conductive skeleton path between key points in the key point set; Based on the conductive skeleton path, the fine-tuning region of the conductive skeleton path is obtained. The pixel architecture of the pixel mesh map is obtained based on the key point set, the conductive skeleton path, and the fine-tuning region. Perform the path search and fine-tuning region operations described above on each pixel grid map of the individual to obtain the pixel structure of each pixel grid map of the individual, and thus obtain the individual; Perform the pixel mesh graph processing described above on each of the key point lists to obtain N individuals, thereby obtaining the initial population. The pixel mesh graph processing is a process of obtaining the individual based on the pixel mesh graph of each individual.

3. The design and optimization method for pixelated electromagnetic structures as described in claim 2, characterized in that, The step of performing path search processing on the keypoint set corresponding to the sublist to obtain the conductive skeleton path between keypoints in the keypoint set includes: Taking the first key point in the set of key points as the starting point of the current path segment and the second key point in the set of key points as the ending point of the current path segment, through A The method searches for the current path segment, then takes the second key point in the key point set as the starting point of the next path segment, and the third key point in the key point set as the ending point of the next path segment. This is done through A. The method searches for the next path segment of the current path segment until it reaches the last key point in the key point set as the endpoint. Each path segment is searched segment by segment to obtain the conductive skeleton path, where each path segment is A. The path segment with the minimum total cost function value of the method.

4. The design and optimization method for pixelated electromagnetic structures as described in claim 2, characterized in that, The process of performing performance evaluation on the initial population to obtain the error value of the initial population includes: Each individual in the initial population is encoded to obtain a pixel encoding matrix for each individual; Based on the pixel encoding matrix of each individual, the actual electromagnetic parameters of each individual are obtained; Based on the actual electromagnetic parameters of each individual and the target electromagnetic parameters, the error value of each individual is obtained, and then the error value of the initial population is obtained. Based on the error value of each individual, the fitness of each individual is obtained.

5. The design and optimization method for pixelated electromagnetic structures as described in claim 4, characterized in that, The micro-optimization process for the initial population includes: The individuals in the initial population are sorted in descending order of fitness, and M1 individuals are selected starting from the individual with the highest fitness in the initial population, where M1 is an integer greater than 1; For each of the M1 individuals, the actual electromagnetic parameters of the individual are compared with the target electromagnetic parameters to obtain a comparison result; Based on the comparison results, candidate pixels of the pixel encoding matrix of the individual are determined, and the candidate pixels are subjected to state flipping processing to obtain the micro-optimized individual; Perform the comparison process described above on each of the M1 individuals until the state flipping process of the candidate pixels of the individual is performed, to obtain the M1 micro-optimized individuals.

6. The design and optimization method for pixelated electromagnetic structures as described in claim 5, characterized in that, The step of determining candidate pixels of the pixel encoding matrix of the individual based on the comparison result, and performing state flipping processing on the candidate pixels to obtain the micro-optimized individual includes: If the comparison result is that the imaginary part difference between the actual electromagnetic parameter of the individual and the imaginary part of the target electromagnetic parameter is not within the imaginary part difference threshold range, and the imaginary part of the actual electromagnetic parameter of the individual is greater than the imaginary part of the target electromagnetic parameter, then a first pixel that satisfies the electric field strength condition and is in a metal-free state is searched in the pixel encoding matrix of the individual. The first pixel is then used as a candidate pixel of the individual, and metal is added to the candidate pixel of the individual to make the candidate pixel of the individual in a metal-containing state until the imaginary part difference is within the imaginary part difference threshold range, thus obtaining the micro-optimized individual. If the comparison result indicates that the imaginary part difference is not within the imaginary part difference threshold range, and the imaginary part value of the actual electromagnetic parameter of the individual is less than the imaginary part value of the target electromagnetic parameter, then a second pixel that satisfies the electric field strength condition and is in a metallic state is searched in the pixel encoding matrix of the individual. The second pixel is then used as a candidate pixel of the individual, and the metal is subtracted from the candidate pixel of the individual to make the candidate pixel of the individual in a non-metallic state. This process continues until the imaginary part difference is within the imaginary part difference threshold range, thus obtaining the micro-optimized individual. If the real part difference is not within the real part difference threshold range, and the real part value of the actual electromagnetic parameter of the individual is greater than the real part value of the target electromagnetic parameter, then a third pixel that satisfies the current density condition and is adjacent to the conductive skeleton path and the fine-tuning region of the individual is searched in the pixel encoding matrix of the individual. The third pixel is then used as a candidate pixel of the individual, and metal is added to the candidate pixel of the individual to make the candidate pixel of the individual have a metallic state until the real part difference is within the real part difference threshold range, thus obtaining the micro-optimized individual. If the comparison result is that the real part difference between the real part of the actual electromagnetic parameter of the individual and the real part of the target electromagnetic parameter is not within the real part difference threshold range, and the real part of the actual electromagnetic parameter of the individual is less than the real part of the target electromagnetic parameter, then a fourth pixel that satisfies the current density condition and is adjacent to the conductive skeleton path of the individual is searched in the pixel encoding matrix of the individual. The fourth pixel is then used as a candidate pixel of the individual, and metal is subtracted from the candidate pixel of the individual to make the candidate pixel of the individual in a metal-free state. This process continues until the real part difference is within the real part difference threshold range, thus obtaining the micro-optimized individual.

7. The design and optimization method for pixelated electromagnetic structures as described in claim 5, characterized in that, The macroscopic optimization process for the initial population includes: The M2 individuals in the initial population are selected by roulette wheel selection, where M2 is an integer greater than 1 and N = M1 + M2. Based on the crossover probability, internal crossover processing is performed on M2 individuals to obtain M2 crossover individuals. The internal crossover processing is a process of exchanging key points in the pixel encoding matrix of the individuals that meet the preset cutting point conditions. Based on the mutation probability, the M2 individuals after the crossover are subjected to mutation processing to obtain M2 mutated individuals, and then M2 macro-optimized individuals are obtained. The mutation processing includes: key point addition processing, key point deletion processing, or key point movement processing. The process of obtaining the current population includes: The current population is obtained by combining the M1 micro-optimized individuals and the M2 macro-optimized individuals.

8. The design and optimization method for pixelated electromagnetic structures as described in claim 4, characterized in that, The error value of the initial population is not within the target error range, including: If the error value of each individual in the initial population is not within the target error range, then it is determined that the error value of the initial population is not within the target error range.

9. An electromagnetic structure, characterized in that, It is formed by the design optimization method of pixelated electromagnetic structure as described in any one of claims 1-8.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the design optimization method for pixelated electromagnetic structures as described in any one of claims 1-8.

Images