Automated design methods for passive components, electronic devices

Abstract

This application provides an automated design method and electronic device for passive devices. The method includes: obtaining the performance optimization target of the passive device; obtaining the geometric parameters of the passive device; using a surrogate model to predict the electromagnetic performance parameters of the passive device based on the geometric parameters; inputting the electromagnetic performance parameters into an optimization algorithm; using the optimization algorithm to optimize the geometric parameters based on the electromagnetic performance parameters and the performance optimization target; and using the optimization algorithm and the surrogate model to iterate the optimization until a preset stopping condition is reached, at which point the iteration stops, and the target geometric parameters are obtained. This application can effectively improve the design efficiency of passive devices.

Description

Technical Field

[0001] This application relates to the field of computer technology, and in particular to an automated design method for passive devices and electronic devices. Background Technology

[0002] In microwave / millimeter-wave RF chip design, the design of passive components (such as inductors and transformers) is a complex process. In the field of millimeter-wave RF chip design, passive components refer to electronic elements that do not possess active energy amplification or control functions, but achieve signal transmission, conversion, filtering, and matching functions solely through their own physical characteristics. These devices do not rely on external power supplies; their operating principle is based on physical phenomena such as electromagnetic induction, electric field coupling, and energy storage, making them core fundamental components for signal processing and energy distribution in RF circuits.

[0003] In related technologies, adjustments to passive components are typically made by experienced engineers. These engineers modify the component's shape and electromagnetic characteristics to meet the performance requirements of the entire circuit. However, when adjusting the electromagnetic characteristics to meet the key parameter requirements of the passive component, its parasitic characteristics also change. Engineers then need to return to the schematic and modify it again to compensate for the effects of these parasitic characteristics, leading to low design efficiency for passive components. Summary of the Invention

[0004] This application provides an automated design method, apparatus, electronic device, computer-readable storage medium, and computer program product for passive devices, which can effectively improve the design efficiency of passive devices.

[0005] The technical solution of this application embodiment is implemented as follows:

[0006] This application provides an automated design method for passive devices, including:

[0007] To obtain the performance optimization target for passive devices;

[0008] Obtain the geometric parameters of the passive device;

[0009] Using a surrogate model, the electromagnetic performance parameters of the passive device are predicted based on the geometric parameters;

[0010] The electromagnetic performance parameters are input into the optimization algorithm, and the geometric parameters are optimized based on the electromagnetic performance parameters and the performance optimization target using the optimization algorithm.

[0011] Using the optimization algorithm and the surrogate model, the optimization is iterated until a preset stopping condition is reached, at which point the iteration stops and the target geometric parameters are obtained.

[0012] This application provides an automated design apparatus for passive devices, comprising:

[0013] The module is used to obtain the performance optimization target of the passive device and the geometric parameters of the passive device.

[0014] The prediction module is used to predict the electromagnetic performance parameters of the passive device based on the geometric parameters using a surrogate model.

[0015] An optimization module is used to input the electromagnetic performance parameters into an optimization algorithm, and use the optimization algorithm to optimize the geometric parameters based on the electromagnetic performance parameters and the performance optimization target;

[0016] The iteration module is used to iterate the optimization using the optimization algorithm and the surrogate model until a preset stopping condition is reached, at which point the iteration stops and the target geometric parameters are obtained.

[0017] This application provides an electronic device, including:

[0018] Memory is used to store executable instructions or computer programs.

[0019] The processor, when executing computer-executable instructions or computer programs stored in the memory, implements the automated design method for passive devices provided in the embodiments of this application.

[0020] This application provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the automated design method for passive devices provided in this application.

[0021] This application provides a computer program product comprising a computer program or computer-executable instructions stored in a computer-readable storage medium. A processor of an electronic device reads the computer-executable instructions from the computer-readable storage medium and executes the computer-executable instructions, causing the electronic device to perform the automated design method for passive devices described in this application.

[0022] The embodiments of this application have the following beneficial effects:

[0023] By obtaining the performance optimization target of passive devices and their geometric parameters, clear input conditions and optimization directions are provided for the entire optimization process.

[0024] By using surrogate models, electromagnetic performance parameters of passive devices can be predicted based on geometric parameters, replacing time-consuming physical testing or electromagnetic simulation, and enabling rapid prediction of electromagnetic performance parameters.

[0025] By inputting electromagnetic performance parameters into the optimization algorithm, the algorithm optimizes the geometric parameters based on the electromagnetic performance parameters and performance optimization objectives. The optimization algorithm automatically adjusts the geometric parameters based on the performance objectives, eliminating the need for manual verification of parameter combinations one by one, thus improving optimization efficiency.

[0026] By utilizing optimization algorithms and surrogate models, the optimization process is iterated until a preset stopping condition is met, at which point the iteration stops and the target geometric parameters are obtained, thus achieving automatic control of the optimization process.

[0027] The entire process eliminates the need for physical fabrication or EM simulation of passive components, reducing the time required in traditional design from several days to just hours or even minutes, effectively improving the design efficiency of passive components. Attached Figure Description

[0028] Figure 1 This is a schematic diagram of the layout of an inductor provided in an embodiment of this application;

[0029] Figure 2 This is a flowchart illustrating the automated design method for passive devices provided in Embodiment 1 of this application;

[0030] Figure 3 This is a flowchart illustrating the automated design method for passive devices provided in Embodiment 2 of this application;

[0031] Figure 4 This is a schematic block diagram of the automated design device provided in Embodiment 2 of this application;

[0032] Figure 5 This is a schematic block diagram of the automated design device provided in Embodiment 3 of this application;

[0033] Figure 6 This is a schematic diagram of the structure of an electronic device for automated design of passive devices provided in an embodiment of this application. Detailed Implementation

[0034] To make the objectives, technical solutions, and advantages of this application clearer, the application will be further described in detail below with reference to the accompanying drawings. The described embodiments should not be regarded as limitations on this application. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0035] In the following description, references are made to “some embodiments,” which describe a subset of all possible embodiments. However, it is understood that “some embodiments” may be the same subset or different subsets of all possible embodiments and may be combined with each other without conflict.

[0036] Passive devices are components that do not require an external power supply and do not contain a voltage or current source. Types of passive devices include transformers, couplers, and inductors. For ease of explanation, the following will use an inductor as an example of a passive device. As an example, Figure 1 This is a schematic diagram of an inductor provided in an embodiment of this application. For example... Figure 1 As shown, the inductor 10 includes at least one coil 11 made of a conductive material, and the geometry of the inductor 10 includes one or more of the following key parameters:

[0037] Inner diameter R: refers to the diameter of the area enclosed by the innermost turn of the coil;

[0038] Line width K: refers to the width of the conductive material that makes up the coil;

[0039] Spacing D1: refers to the distance between two adjacent turns of the coil;

[0040] Number of turns: refers to the total number of turns of a coil continuously wound around its base. Figure 1 The inductor 10 shown has two turns.

[0041] These geometric parameters directly determine the magnetic field distribution, current path, and parasitic effects inside the inductor, thus affecting its electromagnetic performance parameters.

[0042] Electromagnetic performance parameters are quantitative indicators that characterize the functional characteristics and performance of passive devices under specific operating conditions through their internal electromagnetic field distribution, signal transmission, or energy conversion processes. These parameters directly reflect whether the passive device meets the preset performance optimization goals. For example, electromagnetic performance parameters may include at least one of the following: network parameters or noise figure; wherein, network parameters include S-parameters (scattering parameters), Z-parameters (impedance parameters), and / or Y-parameters (admittance parameters). In some embodiments, electromagnetic performance parameters are parameters that can be simulated by electromagnetic simulation (EM simulation).

[0043] S-parameters, or scattering parameters, are parameters that describe the amplitude and phase relationship between incident, reflected, and transmitted waves at each port of a passive device, reflecting the scattering characteristics of a signal within the device. For example, the S-parameters of a two-port inductor 10 can be expressed as:

[0044]

[0045] Where: S11 is the reflection coefficient of port 1, S12 is the transmission coefficient from port 2 to port 1, S21 is the transmission coefficient from port 1 to port 2, and S22 is the reflection coefficient of port 2.

[0046] Z-parameters, or impedance parameters, describe the relationship between voltage and current at each port of a passive device, reflecting the port impedance characteristics of the device. For example, the Z11 parameter of a two-port inductor 10 represents the input impedance of port 1 when port 2 is open. This value needs to match the impedance of the preceding and following circuits (usually 50 ohms), otherwise signal reflection will occur. The Z12 parameter represents the ratio of the voltage at port 1 to the current at port 2 when port 2 is open, reflecting the reverse transmission impedance between the ports. The Z21 parameter represents the ratio of the voltage at port 2 to the current at port 1 when port 1 is open, reflecting the forward transmission impedance between the ports. The Z22 parameter represents the output impedance of port 2 when port 1 is open.

[0047] The Y-parameter, or admittance parameter, is a network parameter that is the reciprocal of the Z-parameter. It describes the relationship between the current and voltage at each port of a passive device, reflecting the admittance characteristics of the device's ports. For example, the Y11 parameter of a two-port inductor 10 characterizes the input admittance of port 1 when port 2 is short-circuited. Admittance is inversely related to impedance, and the Y11 value must match the admittance requirements of the circuit system. The Y12 parameter characterizes the ratio of the current at port 1 to the voltage at port 2 when port 2 is short-circuited, reflecting the reverse transmission admittance between the ports. The Y21 parameter characterizes the ratio of the current at port 2 to the voltage at port 1 when port 1 is short-circuited, reflecting the forward transmission admittance between the ports. The Y22 parameter characterizes the output admittance of port 2 when port 1 is short-circuited.

[0048] The design of passive devices often relies on engineers' experience, involving repeated adjustments to the device's geometric parameters and time-consuming electromagnetic field simulations (EM simulations) to verify electromagnetic performance parameters—a process that is inefficient. Therefore, this application provides an automated design method, apparatus, electronic device, computer-readable storage medium, and computer program product for passive devices, which can effectively improve the design efficiency of passive devices. The exemplary application of the automated design system for passive devices provided in this application is described below.

[0049] To facilitate understanding, the following explanation uses an inductor as an example to illustrate the automated design method for passive components in this solution. See [link to documentation]. Figure 2 , Figure 2 This is a flowchart illustrating the automated design method for passive devices provided in Embodiment 1 of this application.

[0050] First, in step 101, the performance optimization target of the passive device is obtained.

[0051] Performance optimization targets refer to the desired performance indicators set for passive devices during the design process. These include the performance parameters to be optimized and the corresponding target values. When a passive device includes multiple performance optimization targets, weighting coefficients between these targets can be obtained in step 101. These weighting coefficients are used to balance the optimization priorities of different performance optimization targets, providing a quantitative basis for subsequent optimization algorithms to determine the optimization direction. If the electromagnetic performance parameters of the passive device are frequency-dependent, its operating frequency can be specified in step 101. This operating frequency will serve as the input condition for subsequent surrogate model prediction and electromagnetic simulation processes, ensuring that evaluations are performed at this operating frequency.

[0052] The performance optimization targets of an inductor may include at least one of the following: inductance, which characterizes its energy storage capacity; quality factor, which reflects the degree of energy loss; or noise figure. As an example, in this scheme, the target value of inductance can be any value between 20 kHz and 40 kHz, the target value of quality factor can be any value between 10 and 20, the operating frequency can be any value between 40 GHz and 80 GHz, the weighting coefficient of inductance is any value between 1e10 and 1, and the weighting coefficient of quality factor is any value between 1e10 and 1.

[0053] In some embodiments, users can input the type of passive device and its corresponding performance optimization target through an interactive interface or by importing data files, according to design requirements.

[0054] In step 102, the geometric parameters of the passive device are obtained.

[0055] Geometric parameters are key quantitative indicators used to define the physical structure and dimensions of passive devices. They determine the values ​​of the performance parameters of passive devices. For example, the geometric parameters of an inductor can be any one or more of the following: inner diameter, line width, spacing, or number of turns.

[0056] In some embodiments, geometric parameters can be obtained by receiving user input, which can then serve as the starting point for subsequent optimization processes. Specifically, users can provide initial geometric parameters related to the passive device structure based on their design experience or preliminary plans through an interactive interface or data import function.

[0057] In step 103, the electromagnetic performance parameters of the passive device are predicted based on the geometric parameters using a surrogate model.

[0058] A surrogate model is a mathematical model trained on historical data, which includes multiple known correspondences between geometric parameters and electromagnetic performance parameters. These correspondences can be obtained through prior electromagnetic simulation or experimental measurements. Multiple geometric parameters can be input into an electromagnetic (EM) simulator to obtain corresponding electromagnetic performance parameters, thus generating multiple sets of geometric parameter-electromagnetic performance parameter correspondences. The geometric parameters and their corresponding electromagnetic performance parameters are then input into the surrogate model for training. The surrogate model learns the mapping patterns in the historical data. When new geometric parameters are input into the surrogate model, it can predict the corresponding electromagnetic performance parameters at a speed faster than electromagnetic simulation.

[0059] In this context, the geometric parameters during the model training phase refer to the set of geometric parameter samples selected within a reasonable range of values ​​for the geometric parameters of the target passive device in order to construct the surrogate model. The new geometric parameters refer to the geometric parameters received by the surrogate model in step 103 or step 105.

[0060] In this embodiment, the electromagnetic performance parameters are S-parameters, and the surrogate model can predict S11, S12, S21, and S22.

[0061] In step 104, the electromagnetic performance parameters are input into the optimization algorithm, and the geometric parameters are optimized based on the electromagnetic performance parameters and the performance optimization target.

[0062] An optimization algorithm is a computational method for searching for optimal solutions. In some embodiments, the optimization algorithm receives electromagnetic performance parameters predicted by a surrogate model and determines the optimization direction of geometric parameters based on the deviation between the performance corresponding to the electromagnetic performance parameters and the performance optimization target.

[0063] Specifically, after the electromagnetic performance parameters (S-parameters) predicted by the surrogate model are input into the optimization algorithm, the optimization algorithm will calculate the corresponding inductance L1 and quality factor Q1 based on the S-parameters (S11, S12, S21, S22).

[0064] The formulas for calculating inductance L and quality factor Q are as follows:

[0065] z=z0×(1+(S11-S12×S21 / (1+S22)))×(1-(S11-S12×S21 / (1+S22)))

[0066]

[0067] Q = imag(z) / real(z)

[0068] Where z0 is the port impedance (z0 is usually 50 ohms), z is the calculated impedance, freq is the operating frequency of the inductor, pi is (constant), imag(z) is the imaginary part of z, and real(z) is the real part of z.

[0069] For example, based on the formulas for calculating inductance L and quality factor Q, we can calculate L1 = 200 picometers and Q1 = 22 picometers.

[0070] The optimization algorithm calculates the deviations between the inductance L1 and the quality factor Q1 and the preset performance optimization targets (e.g., the target value of inductance Lt is 30 picofarads, and the target value of quality factor Qt is 20), for example, by constructing a loss function to quantify the deviations:

[0071] The loss function is ((L1-Lt)×W_L) 2 +((Q1-Qt)×W_Q) 2 ,

[0072] The weights of inductance L and quality factor Q are W_L=1e10 and W_Q=1, respectively, resulting in a loss function of 5. The loss function can be represented by "loss" or "fitness".

[0073] The optimization algorithm analyzes the direction and extent of the deviation, and then determines the direction and magnitude of adjustment for the geometric parameters of passive components (such as the inner diameter, number of turns, line width, and spacing of inductors). For example, if the inductance in the electromagnetic performance parameters is lower than the optimization target, the algorithm will determine that the number of turns of the inductor or the inner diameter needs to be adjusted to increase the inductance.

[0074] In step 105, the optimization algorithm and surrogate model are used to iterate the optimization until a preset stopping condition is reached, at which point the iteration stops and the target geometric parameters are obtained.

[0075] The target geometric parameters refer to the geometric parameters that are finally determined and output when the optimization iteration process meets the preset stopping conditions, and that meet the performance optimization objectives.

[0076] In some embodiments, the iterative process repeatedly executes the steps of "optimization algorithm generates new geometric parameters → surrogate model predicts electromagnetic performance parameters → optimization algorithm evaluates electromagnetic performance parameters and adjusts geometric parameters". Using the optimization algorithm and surrogate model, through multiple rounds of geometric parameter adjustment and electromagnetic performance parameter prediction, steps 103 and 104 are repeated to iteratively optimize the geometric parameters of the inductor, ensuring that its corresponding electromagnetic performance parameters approach or meet the performance optimization target. Iteration stops when a preset stopping condition is met, ultimately obtaining the target geometric parameters and ensuring that the final target geometric parameters meet the circuit application requirements.

[0077] Steps 101 to 105 achieve an efficient automated design process:

[0078] By obtaining the performance optimization target of passive devices and their geometric parameters, clear input conditions and optimization directions are provided for the entire optimization process.

[0079] By using surrogate models, electromagnetic performance parameters of passive devices can be predicted based on geometric parameters, replacing time-consuming physical testing or electromagnetic simulation, and electromagnetic performance parameters can be predicted quickly.

[0080] By inputting electromagnetic performance parameters into the optimization algorithm, the algorithm optimizes the geometric parameters based on the electromagnetic performance parameters and performance optimization objectives. The optimization algorithm automatically adjusts the geometric parameters based on the performance objectives, eliminating the need for manual verification of parameter combinations one by one, thus improving optimization efficiency.

[0081] By utilizing optimization algorithms and surrogate models, the optimization is iterated until a preset stopping condition is reached, at which point the iteration stops and the target geometric parameters are obtained. Based on the target geometric parameters, physical passive devices can be manufactured, thus realizing the automatic control of the optimization process.

[0082] The entire process does not require physical fabrication or EM simulation of passive components, reducing the time that may take several days in traditional design to a few hours or even minutes, effectively improving the design efficiency of passive components.

[0083] In terms of design effectiveness, this method, through the synergy of a surrogate model and an optimization algorithm, can discover passive circuit designs that outperform traditional design experience, for example:

[0084] In the inductor design of low-noise amplifiers, this method optimizes and obtains a high-performance inductor coil structure, which successfully replaces the original three simple coils and significantly reduces the chip footprint.

[0085] In the design of passive balun circuits in broadband impedance matching circuits, applying this method increases the bandwidth of the balun circuit by approximately 50%.

[0086] In the passive transformer design of multistage amplifiers, this method optimizes and reduces the number of transformers used for compensation, resulting in a reduction of the layout area by approximately 40%.

[0087] In the design of passive couplers in phase shifters, this method optimizes a single coupler to replace the original two couplers, achieving a reduction of approximately 50% in layout area while also reducing insertion loss by approximately 0.5 dB.

[0088] In some embodiments, the preset stopping conditions include multiple types, and a single condition or a combination of conditions can be selected according to optimization needs:

[0089] The first method involves taking the geometric parameters of the last iteration as the target geometric parameters when the number of iterations reaches a preset number (e.g., 50-100 cycles), ensuring that the iteration is completed within a controllable time.

[0090] The second scenario involves multiple iterations where the deviation between the electromagnetic performance parameters predicted by the surrogate model and the performance optimization target is less than a preset threshold (e.g., the value of the loss function is less than the preset threshold of 0.5). This indicates that the electromagnetic performance parameters have approached a stable value, and further iterations offer limited effective adjustment of the parameters. In this case, the iteration is stopped, and the geometric parameters from the last round are used as the target geometric parameters. The preset threshold is a threshold value set by the user, and the "number of consecutive iterations" can be set to 5-8.

[0091] The third type is when, in multiple iterations (such as 5-10 rounds), the electromagnetic performance parameters corresponding to the optimal geometric parameters do not improve significantly. For example, if the change in the loss function is less than or equal to the set value, it indicates that the geometric parameters have approached the optimal solution and the optimization has converged. The iteration can be stopped, and the geometric parameters of the last round can be used as the target geometric parameters.

[0092] In some embodiments, when the iteration reaches any or a combination of the above stopping conditions, the iteration process is stopped, and the geometric parameters that best match the performance optimization objective predicted by the surrogate model from the final round of geometric parameter combinations are selected as the target geometric parameters; or, the geometric parameters optimized in the last iteration are selected as the target geometric parameters. For example, when the iteration reaches the 60th round, if the deviations corresponding to a certain set of geometric parameters are all less than a preset threshold, then the geometric parameters are the final target geometric parameters.

[0093] In some embodiments, the execution order of steps 101 and 102 is not limited. Step 101 can be executed first to obtain the performance optimization target, and then step 102 can be executed to obtain the geometric parameters; or step 102 can be executed first to obtain the geometric parameters, and then step 101 can be executed to obtain the performance optimization target; or steps 101 and 102 can be executed simultaneously to obtain the performance optimization target and geometric parameters in parallel.

[0094] In some embodiments, obtaining the geometric parameters of the passive device in step 102 may include:

[0095] Step 1021: Obtain the device type of the passive device to determine the specific structural category and functional attributes of the passive device to be optimized. Different device types correspond to different structural features and geometric parameters. Users can input the device type through an interactive interface or by importing data files. For example, the user can specify the device type as "regular octagonal inductor".

[0096] Step 1022: Using an optimization algorithm, initialize the geometric parameters according to the device type and performance optimization objective. Initializing the geometric parameters refers to the geometric parameters provided by the optimization algorithm to the surrogate model for the first prediction at the start of the optimization iteration. The geometric parameters can be initialized in any of the following ways:

[0097] Method 1: The optimization algorithm can randomly select from a preset spatial sample to generate initial geometric parameters. Specifically, the optimization algorithm first determines the reasonable value range of each geometric parameter based on the type of passive device to be optimized (e.g., a regular octagonal inductor) and performance optimization objectives (e.g., inductance of 300 picohens, quality factor of 20, operating frequency of 60 gigahertz), combined with process limits (e.g., minimum linewidth, maximum number of turns), design experience of similar devices, and the sensitivity of performance to each parameter. These values ​​collectively constitute the preset spatial sample. For example, for a regular octagonal inductor, the value range of its octagonal inner diameter can be set to 10 micrometers to 20 micrometers, the value range of the number of coil turns can be set to 1 to 5 turns, the value range of the linewidth can be set to 2 micrometers to 4 micrometers, and the value range of the spacing can be set to 1 micrometer to 2 micrometers. The value ranges of each geometric parameter collectively constitute the spatial sample. Subsequently, a random selection algorithm is used to randomly select a set of numerical combinations within the range of parameter dimensions in this spatial sample. These combinations are then used as the initial geometric parameters. For example, 18 micrometers can be randomly selected from the inner diameter range, 2 turns from the number of turns range, 3 micrometers from the linewidth range, and 1 micrometer from the spacing range. These combinations form the initial geometric parameters for the regular octagonal inductor. This method can automatically provide relatively accurate initial geometric parameters for subsequent optimization.

[0098] Method Two: The optimization algorithm can generate fixed geometric parameters based on preset geometric parameters, which are then used as initialization parameters. These preset geometric parameter combinations are a set of validated and reasonable geometric parameters determined based on device type, performance optimization goals, and historical design data. For example, for a regular octagonal inductor with a 60 GHz frequency band, 300 picohen inductance, and a quality factor of 20, the preset geometric parameters can be set as follows: octagonal inner diameter 10 micrometers, coil turns 2 turns, linewidth 4 micrometers, and spacing 2 micrometers. During initialization, these preset geometric parameters are directly used as initial values, eliminating the need for additional random selection or adjustment. This ensures high rationality and stability of the initial geometric parameters, reducing the increase in iterations caused by deviations from reasonable ranges in subsequent optimization processes. This method is particularly suitable for scenarios with high optimization efficiency requirements or those with existing mature design foundations.

[0099] In some embodiments, the optimization algorithm includes a genetic algorithm or a simulated annealing algorithm. Both the genetic algorithm and the simulated annealing algorithm can output optimized geometric parameters based on a loss function, which can reduce the dependence on human experience and realize the automation and efficiency of the passive device design process.

[0100] Genetic algorithms simulate the evolutionary process of biological populations, searching for the optimal solution (optimal geometric parameters) through multiple generations of iteration. Specifically, this includes:

[0101] Population initialization: Within the preset range of geometric parameters, a set of initial solutions is randomly generated. Each initial solution constitutes an "individual", and all individuals together form the initial "population".

[0102] Calculate the fitness value: Based on the loss function, calculate the fitness value. The smaller the loss function, the higher the corresponding fitness value.

[0103] Selection operation: Based on the fitness value of an individual, select individuals with higher fitness values ​​from the current population as "parents" using a roulette wheel or tournament method, and retain the characteristics of high-quality parameter combinations.

[0104] Crossover operation: Selected parent individuals are paired according to a preset probability, and some of their geometric parameters are exchanged to generate new "offspring" individuals, thereby realizing the recombination and optimization of parameter combinations.

[0105] Mutation operation: Randomly change the values ​​of one or more geometric parameters in offspring individuals with a low probability, introduce new parameter combinations, and avoid the algorithm getting trapped in local optima.

[0106] Iterative update: The newly generated offspring individuals are merged with the parent individuals, and the individuals with the highest fitness values ​​are selected to form a new generation of population. The above operations of calculating fitness values, selection, crossover, and mutation are repeated until the preset stopping condition is met.

[0107] Output: When the iteration ends, select the geometric parameters corresponding to the individual with the highest fitness value from the final population as the target geometric parameters.

[0108] Simulated annealing is used to simulate temperature-induced annealing. By controlling the "temperature" parameter during the search process, it balances global exploration with local exploitation to avoid getting trapped in local optima. The "temperature" parameter controls the search process of the optimization algorithm. In the early stages of iteration, the temperature is set to a relatively high initial value, and as iterations progress, the temperature monotonically decreases according to a predetermined cooling plan. The simulated annealing algorithm specifically includes:

[0109] Initialization: Set the initial temperature, initial solution, and temperature drop coefficient. The initial solution is a set of initial geometric parameters randomly selected based on the design range of the geometric parameters.

[0110] Calculate the energy value: Based on the loss function, calculate the corresponding energy value. The larger the loss function, the higher the energy value.

[0111] Generate neighborhood solutions: Based on the current solution, randomly adjust one or more geometric parameters according to a preset step size to generate new candidate solutions.

[0112] Determining whether to accept: Calculate the energy value of the neighborhood solution. If the energy value of the neighborhood solution is lower than that of the current solution, then directly accept the neighborhood solution as the new current solution. If the energy value of the neighborhood solution is higher than that of the current solution, calculate the acceptance probability according to the Metropolis criterion, and use this probability to decide whether to accept the neighborhood solution, thus preserving a limited capacity to accept worse solutions.

[0113] Cooling: Reduce the current temperature according to the preset temperature reduction coefficient.

[0114] Iterative update: Repeat the above operations of calculating energy value, generating neighborhood solution, determining whether to accept and cooling, until the preset stopping condition is met.

[0115] Output: When the iteration ends, the geometric parameters corresponding to the current solution obtained in the last iteration are used as the target geometric parameters.

[0116] In the second embodiment of this solution, the proxy model can also be updated.

[0117] like Figure 3 As shown, step 103, while predicting electromagnetic performance parameters, may also include:

[0118] In step 1031, the surrogate model also outputs a statistical discrete metric based on the predicted electromagnetic performance parameters.

[0119] Statistical dispersion measures are used to characterize the confidence level of a prediction. When a surrogate model makes predictions based on the geometric parameters of passive devices, it is a statistical indicator output synchronously by the surrogate model to quantify the dispersion and uncertainty of the predicted results. It reflects the reliability of the prediction results numerically, thus providing a basis for judging the confidence level of the prediction results.

[0120] The essence of this metric is to assess the range by which the predicted results may deviate from the true values ​​by analyzing the correspondence between "geometric parameters and electromagnetic performance parameters" relied upon in the construction of the surrogate model. The magnitude of the statistical discreteness metric is positively correlated with the uncertainty of the prediction: the lower the value, the more concentrated the possible distribution of the predicted values, the higher the certainty, and the more reliable the prediction results; the higher the value, the more dispersed the possible distribution of the predicted values, the greater the uncertainty, and the lower the reliability of the prediction results.

[0121] In this embodiment, the statistical dispersion measure includes variance (σ²) or standard deviation (σ). For example, for a regular octagonal inductor, if the standard deviation of the surrogate model output is 0.5, it indicates that the predicted inductance value is less dispersed near the true value, and the prediction result is highly reliable; if the standard deviation is 5, it indicates that the predicted inductance value is more dispersed, and the prediction result is less reliable.

[0122] See Figure 3 Furthermore, this application embodiment also uses statistical discreteness measures to update the surrogate model, thereby continuously improving the accuracy and reliability of model predictions. Specifically, the automated design method may also include:

[0123] Step 201: When the statistical discrete metric is greater than the set index in the current iteration, perform electromagnetic simulation on the current geometric parameters to obtain the simulated electromagnetic performance parameters.

[0124] The set index is a pre-configured value used to judge the reliability of the prediction result. This index is compared with the statistical dispersion measure (such as variance or standard deviation) output by the surrogate model. If the statistical dispersion measure is less than or equal to the set index, it indicates that the confidence level of the prediction result meets the requirements. Conversely, when the statistical dispersion measure is greater than the set index, it indicates that the current prediction uncertainty of the electromagnetic performance parameters is high and the confidence level is insufficient; if geometric parameter optimization is continued based on this prediction result, it may cause the optimization direction to deviate from the actual performance requirements of passive devices. The set index can be any positive real number, and the value of the set index needs to be set according to the actual situation to balance the speed and accuracy of the optimization process. The higher the value of the set index, the faster the optimization process; the lower the value of the set index, the higher the accuracy of the optimization process. If the value is too low, it may lead to a longer optimization process. Generally, the value of the set index can be 0.00001~0.1.

[0125] When the statistical discrete metric output by the surrogate model for the current geometric parameters is greater than the preset index during the current iteration, electromagnetic simulation is performed on the passive device corresponding to the current geometric parameters to simulate the electromagnetic field distribution and signal transmission law under preset operating conditions, and finally output simulated electromagnetic performance parameters that can reflect the true performance of the device.

[0126] In some embodiments, geometric parameters can be first input into drawing software (such as drawing scripts) to convert the geometric parameters into an inductor layout, and then the layout can be input into an EM simulator for electromagnetic simulation to obtain the simulated electromagnetic performance parameters.

[0127] Step 202: Update the surrogate model based at least on the electromagnetic performance parameters from the last simulation and the current geometric parameters.

[0128] The current geometric parameters refer to the geometric parameters that are temporarily received by the surrogate model and used for prediction in a certain iteration.

[0129] After obtaining the electromagnetic performance parameters of the simulation, these parameters and the current geometric parameters are added as new sample data to the training dataset of the surrogate model. At least the electromagnetic performance parameters of the last simulation and the current geometric parameters are used as supplementary data. Combined with the original "geometric parameters-electromagnetic performance parameters" sample data of the model, the surrogate model is retrained and adjusted to update the surrogate model.

[0130] This update method can supplement high-reliability real simulation data at high uncertainty prediction points where statistical discreteness exceeds the standard, optimize the training of the model on the performance mapping relationship corresponding to this type of geometric parameter combination, and make the updated surrogate model more accurate in predicting the electromagnetic performance parameters of similar geometric parameters in subsequent iterations. This improves the reliability and efficiency of the entire geometric parameter optimization process and ensures that the final target geometric parameters can effectively meet the performance optimization goals of passive devices.

[0131] By introducing a statistical discrete metric and an automatic update mechanism for the surrogate model, both the accuracy and time of optimization are balanced:

[0132] The judgment module determines whether the statistical discrete measure exceeds the set index. When the statistical discrete measure exceeds the set index, the judgment module automatically triggers electromagnetic simulation and replaces the low confidence prediction value with high-precision simulation data. This effectively avoids the problem of optimization deviating from actual needs due to the prediction deviation of the surrogate model and ensures the accuracy of the optimization direction.

[0133] This mechanism decouples electromagnetic simulation from the optimization process by selectively triggering electromagnetic simulation, eliminating the need for electromagnetic simulation in every iteration and thus reducing automated design time.

[0134] In some embodiments, the EM simulation takes a long time. By performing steps 201 and 202 asynchronously with steps 104 and 105, parallel execution of optimization iteration and model update is achieved. Specifically, while executing steps 201 and S202 (i.e., EM simulation and GPR model update), steps 104 and S105 (i.e., optimization iteration based on the current GPR model) are performed asynchronously. This parallel architecture ensures that the time-consuming EM simulation does not block the main iteration flow of the optimization algorithm, thereby improving design efficiency.

[0135] In some embodiments, the preset stopping condition in step 105 may also include: the statistical discrete measure is less than or equal to a set index, and the statistical discrete measure output by the surrogate model is used as one of the judgment criteria to ensure that the geometric parameters at the time of iteration stop not only fit the performance optimization target, but also that the confidence of the prediction result meets the requirements.

[0136] Specifically, the "statistical dispersion measure is less than or equal to the set index" is used in combination with stopping conditions such as the change in geometric parameters being less than a preset threshold, the number of iterations reaching a preset number, and / or the deviation not converging, to jointly determine whether to stop the iteration. When all stopping conditions are met, the iteration process terminates, and the optimal geometric parameters of the current round are determined as the target geometric parameters.

[0137] In some embodiments, the surrogate model is a Gaussian process regression (GPR) model, which is a nonlinear regression model based on probability and statistics theory, including a mean function m(x) and a kernel function. The mean function m(x) represents the predicted value when there is no data, while the kernel function represents the smoothness and rate of change of the function value. The kernel function, also known as the covariance function, includes variance or standard deviation and is used to characterize a measure of statistical dispersion.

[0138] In some embodiments, the kernel function is:

[0139]

[0140] in, is the variance; L is the length factor, used to quantify the decay rate of the influence of the input geometric parameters on the output electromagnetic performance parameters; x is the geometric parameter in the historical samples, and x' is the geometric parameter input in step 103.

[0141] The GPR model describes the mapping relationship between input variables (geometric parameters of passive devices) and output variables (electromagnetic performance parameters) through a pre-defined kernel function, and uses the probabilistic characteristics of Gaussian processes to predict the output results. At the same time, it quantifies the uncertainty of the prediction results, which meets the needs of predicting electromagnetic performance parameters and outputting statistical discrete metrics mentioned above.

[0142] In the process of optimizing the geometric parameters of an inductor, when constructing the GPR model, the sample data of geometric parameters and electromagnetic performance parameters acquired in history are first input into the model. The correlation characteristics between different combinations of geometric parameters and corresponding electromagnetic performance parameters are learned through kernel functions (such as square exponential kernel, Marton kernel, etc.) to determine the mean function and kernel function of the Gaussian process.

[0143] In step 103, when predicting electromagnetic performance parameters, the current geometric parameters to be evaluated are input into the trained GPR model. The model will output the predicted value of the electromagnetic performance parameters corresponding to the geometric parameters based on the learned correlation features, and at the same time, the variance or standard deviation of the prediction results is calculated through the kernel function.

[0144] When updating the surrogate model, simply add newly acquired electromagnetic simulation data (such as simulated electromagnetic performance parameters obtained when the statistical discrete metric exceeds the limit in the current iteration) as new samples to the training dataset to train the surrogate model; and / or, based on the training results, replace the hyperparameters (such as length factors) of the mean function and / or kernel function to minimize the model's fitting error to the training set data. Simultaneously, adjust the model parameters using validation set data to complete the training and calibration of the surrogate model. Updating the surrogate model does not require significant adjustments to the model structure. The update process is simple and effectively preserves the learning results of the original samples, continuously improving the model's fitting accuracy to the mapping relationship between geometric parameters and electromagnetic performance parameters. This provides a more reliable performance prediction basis for subsequent optimization algorithms to adjust geometric parameters, ensuring the entire optimization process is efficient and accurate.

[0145] Figure 4 This is a schematic block diagram of the automated design device provided in Embodiment 2 of this application. The device includes an optimization algorithm, a GPR model, a judgment module, a drawing script, and an EM simulator. The following is combined with... Figure 3 and Figure 4 The working process of this device is explained below:

[0146] In step 101, the user inputs the circuit structure and optimization objective into the optimization algorithm. The circuit structure is a passive inductor, and the performance optimization objective is an inductance of L=300 picometers, a quality factor of Q=20, an operating frequency of 60 GHz, and weights of inductance and quality factor of W_L=1e10 and W_Q=1, respectively.

[0147] In step 102, the optimization algorithm can use simulated annealing or a genetic algorithm to generate a set of initial solutions as the initial device geometry parameters, such as inner diameter = 20 μm, line width = 5 μm, spacing = 2 μm, and number of turns = 2.

[0148] In step 103, the optimization algorithm inputs the geometric parameters into the GPR model, and the GPR model predicts the electromagnetic performance parameters (e.g., S-parameters) as S1; at the same time, the GPR model outputs the statistical dispersion measure (e.g., standard deviation) corresponding to the S-parameters S1 as 0.1.

[0149] The judgment module compares the standard deviation of 0.1 with a set upper limit (e.g., 0.0001). If the standard deviation of 0.1 is greater than the set upper limit, the judgment module determines that the current prediction uncertainty is high (high standard deviation), and triggers step 201. The judgment module inputs the geometric parameter into the drawing script, which converts the geometric parameter into a layout. The EM simulator performs EM simulation based on the layout. After the EM simulation is completed, the electromagnetic performance parameters (e.g., simulation S-parameters) corresponding to the geometric parameter are obtained. Then, step 202 is performed to update the surrogate model based on the simulation S-parameters and the geometric parameter.

[0150] In step 104, the GPR model inputs the S-parameters into the optimization algorithm. Based on the S-parameters S1 and the formulas for calculating inductance L and quality factor Q, the optimization algorithm obtains L1 = 200p and Q1 = 22 corresponding to S1. The optimization algorithm then compares L1 and Q1 with the performance optimization target and calculates the loss = ((L1 - L) × W_L). 2 +((Q1-Q)×W_Q) 2 =5. The optimization algorithm will generate a new solution based on the loss, for example, inner diameter = 22um, line width = 4.6um, spacing = 2um, number of turns = 2.

[0151] In step 105, the optimization algorithm generates a new solution, inputs it into the updated GPR model, and the updated GPR model calculates the S-parameter as S2. Based on the S-parameter S2, the optimization algorithm recalculates the loss and generates a new solution to iterate until the preset stopping condition is reached.

[0152] In the third embodiment of this solution, as Figure 5 As shown, the surrogate model may also include a pre-trained neural network model and a GPR model, which work together to provide accurate predictions of electromagnetic performance parameters based on the geometric parameters input by the optimization algorithm.

[0153] In some embodiments, a neural network model may include an input layer, hidden layers, and an output layer. Each layer consists of multiple neurons. The operation of each neuron can be represented as... Where x is the input vector (e.g., geometric parameters), w is the weight parameter vector of the neuron, and b is the bias of the neuron. This is the activation function.

[0154] Given that training a neural network model relies on large-scale sample data, this embodiment uses a low-precision electromagnetic simulator to generate training samples to improve training efficiency. The electromagnetic simulation precision of the low-precision electromagnetic simulator is lower than that in step 201, allowing for rapid acquisition of large-scale simulated electromagnetic performance parameters. Specifically, a large number of geometric parameters are rapidly simulated using the low-precision simulator to obtain the corresponding low-precision electromagnetic performance parameters SL. The neural network model is then trained using the "geometric parameters - low-precision electromagnetic performance parameters SL" data. Here, "large number of geometric parameters" refers to the set of geometric parameter samples selected within a reasonable range of values ​​for the geometric parameters of the target passive device during the model training phase, for building the neural network model training dataset. This set is relatively large, for example, greater than 200 samples.

[0155] To compensate for the systematic errors introduced by low-precision simulation and improve the overall prediction accuracy, this embodiment introduces the GPR model to calibrate the errors. A high-precision electromagnetic simulator is used to accurately simulate a portion of the geometric parameters from a large number of geometric parameters, obtaining the high-precision electromagnetic performance parameter SH. The electromagnetic simulation accuracy of the high-precision electromagnetic simulator is equal to the electromagnetic simulation accuracy in step 201. The error between this parameter and the neural network prediction value SL is calculated, i.e., the parameter error ΔS = SH. SL. Subsequently, the GPR model is trained based on "geometric parameters - parameter error ΔS".

[0156] Furthermore, step 1031 includes:

[0157] Using a neural network model, the initial electromagnetic performance parameters, such as the initial S-parameter SL-1, are predicted based on the geometric parameters input to the optimization algorithm.

[0158] Using the GPR model, based on the geometric parameters input by the optimization algorithm, the parameter error (e.g., S-parameter error ΔS1) and the predicted statistical discrete measure (e.g., standard deviation) are predicted. The parameter error is the difference between the initial electromagnetic performance parameter and the electromagnetic performance parameter of the electromagnetic simulation for the same geometric parameter. In this embodiment, the statistical discrete measure indicates the reliability of the predicted parameter error.

[0159] The electromagnetic performance parameters are obtained by summing the initial electromagnetic performance parameters (SL1) and the parameter error (ΔS1), for example, the modified S-parameter S1 = SL1 + ΔS1. Figure 5 As shown, the sum of the initial electromagnetic performance parameters and parameter errors, S = SL + ΔS, can be calculated using the summation module.

[0160] When the judgment module determines that the geometric parameter is high based on the standard deviation, step 201 is triggered. The judgment module inputs the geometric parameter into the drawing script, which converts the geometric parameter into a layout. The EM simulator performs EM simulation based on the layout. After the EM simulation is completed, the simulation S-parameter SH1 corresponding to the geometric parameter is obtained. Then, step 202 is performed. Based on the simulation S-parameter SH1, the parameter error ΔS2 = SH1 - SL1 is calculated. The parameter error ΔS2 and the geometric parameter are input into the GPR model for training to update the GPR model.

[0161] The neural network model is pre-trained with a large amount of low-precision simulation data, which can output relatively accurate initial electromagnetic performance parameters, providing a reliable benchmark. The GPR model is used to learn the error between the neural network prediction results and high-precision simulation data, reducing the simulation time of the GPR model's high-precision training data, while reducing the prediction range of the GPR model, reducing the nonlinearity of the prediction data, and improving the accuracy of the GPR model.

[0162] The above description uses an inductor as an example to illustrate the automated design method of this application. In other embodiments, passive devices may also be transformers, couplers, capacitors, etc.

[0163] When the passive device is a transformer, its performance optimization targets may include inductance and / or quality factor; geometric parameters may include the number of turns of the primary coil, the number of turns of the secondary coil, line width, spacing, coil size and coupling distance between the two coils; electromagnetic performance parameters may include S-parameters, insertion loss, return loss or noise figure calculated based on S-parameters.

[0164] When the passive device is a coupler, its performance optimization objectives may include coupling degree; geometric parameters may include the length, width, and spacing of the main line and the coupled line; electromagnetic performance parameters may include S-parameters and noise figure.

[0165] When the passive device is a capacitor, its performance optimization targets may include capacitance value; geometric parameters may include plate area, plate spacing, dielectric layer thickness and edge structure; electromagnetic performance parameters may include S-parameters, Y-parameters, etc.

[0166] By adapting to different types of passive devices and adjusting the corresponding performance optimization targets, geometric parameters, and electromagnetic performance parameters, the method of this application embodiment can be widely applied to the design optimization of various passive devices.

[0167] See Figure 6 , Figure 6 This is a schematic diagram of the structure of an electronic device 500 for automated design of passive devices provided in an embodiment of this application. Figure 6 The illustrated electronic device 500 includes at least one processor 430 and a memory 450. Various components in the electronic device 500 are coupled together via a bus system 440. It is understood that the bus system 440 is used to implement communication between these components. In addition to a data bus, the bus system 440 also includes a power bus, a control bus, and a status signal bus. However, for clarity, in… Figure 6 The general labeled all buses as Bus System 440.

[0168] Processor 430 may be an integrated circuit chip with signal processing capabilities. Memory 450 may be removable, non-removable, or a combination thereof. Exemplary hardware devices include solid-state storage, hard disk drives, optical disk drives, etc.

[0169] In some embodiments, memory 450 is capable of storing data to support various operations, examples of which include programs, modules, and data structures, or subsets or supersets thereof, as illustrated below. Operating system 451 includes system programs for handling various basic system services and performing hardware-related tasks.

[0170] The following description continues to illustrate the exemplary structure of the automated design device 455 for passive devices provided in this application embodiment as a software module. In some embodiments, such as... Figure 6 As shown, the software modules in the automated design device 455 for passive devices stored in the memory 450 may include: optimization algorithms, surrogate models, judgment modules, drawing scripts, and EM simulators.

[0171] This application provides a computer program product, which includes a computer program or computer-executable instructions stored in a computer-readable storage medium. The processor of an electronic device reads the computer-executable instructions from the computer-readable storage medium and executes the instructions, causing the electronic device to perform the following automated design method: obtaining a performance optimization target for a passive device; obtaining the geometric parameters of the passive device; using a surrogate model to predict the electromagnetic performance parameters of the passive device based on the geometric parameters; inputting the electromagnetic performance parameters into an optimization algorithm; using the optimization algorithm to optimize the geometric parameters based on the electromagnetic performance parameters and the performance optimization target; and using the optimization algorithm and the surrogate model to iterate the optimization until a preset stopping condition is reached, at which point the iteration stops, thus obtaining the passive device with the target geometric parameters.

[0172] This application provides a computer-readable storage medium storing computer-executable instructions. When the computer-executable instructions are executed by a processor, the processor will execute the automated design method for passive devices provided in this application.

[0173] In some embodiments, the computer-readable storage medium may be a memory such as FRAM, ROM, PROM, EPROM, EEPROM, flash memory, magnetic surface memory, optical disk, or CD-ROM; or it may be a variety of electronic devices including one or any combination of the above-mentioned memories.

[0174] In some embodiments, computer-executable instructions may take the form of programs, software, software modules, scripts, or code, written in any form of programming language (including compiled or interpreted languages, or declarative or procedural languages), and may be deployed in any form, including as stand-alone programs or as modules, components, subroutines, or other units suitable for use in a computing environment.

[0175] The above are merely embodiments of this application and are not intended to limit the scope of protection of this application. Any modifications, equivalent substitutions, and improvements made within the spirit and scope of this application are included within the scope of protection of this application.

Claims

1. An automated design method for passive devices, characterized in that, The method comprises the following steps: obtaining a performance optimization target of a passive device; obtaining geometric parameters of the passive device; using a surrogate model to predict electromagnetic performance parameters of the passive device according to the geometric parameters, and outputting a statistical dispersion metric according to the predicted results, the statistical dispersion metric being used to represent a confidence level of the prediction; inputting the electromagnetic performance parameters into an optimization algorithm, and using the optimization algorithm to optimize the geometric parameters based on the electromagnetic performance parameters and the performance optimization target; stopping iteration when a preset stop condition is reached by using the optimization algorithm and the surrogate model to iteratively optimize the optimization, and obtaining target geometric parameters; when the statistical dispersion metric is greater than a set index in the current iteration, performing electromagnetic simulation on the current geometric parameters to obtain simulated electromagnetic performance parameters; and updating the surrogate model according to the simulated electromagnetic performance parameters and the current geometric parameters, the simulated electromagnetic performance parameters replacing the previously predicted electromagnetic performance parameters, wherein the accuracy of the simulated electromagnetic performance parameters is higher than that of the previously predicted electromagnetic performance parameters.

2. The automatic design method of the passive device according to claim 1, wherein the preset stop condition comprises: the statistical dispersion metric being less than or equal to the set index. The surrogate model is a GPR model.

3. The method of claim 1, wherein, The surrogate model comprises a neural network model and a GPR model.

4. The method of claim 1, wherein, 5. The automatic design method of the passive device according to claim 4, wherein the surrogate model predicts the electromagnetic performance parameters and outputs the statistical dispersion metric according to the predicted results, comprising: using the neural network model to predict initial electromagnetic performance parameters according to the current geometric parameters; using the GPR model to predict a parameter error according to the current geometric parameters and output the statistical dispersion metric; the parameter error being a difference between the initial electromagnetic performance parameters and the simulated electromagnetic performance parameters for the same geometric parameters; obtaining the predicted electromagnetic performance parameters according to a sum of the initial electromagnetic performance parameters and the parameter error.

6. The automatic design method of the passive device according to claim 1, wherein the surrogate model comprises a mean function and a kernel function; and the updating of the surrogate model comprises: replacing a predicted mean of the mean function and / or a hyperparameter of the kernel function of the surrogate model, and / or training the surrogate model by taking the geometric parameters and the electromagnetic performance parameters as new samples. The statistical dispersion metric comprises a variance or a standard deviation.

8. The automatic design method of the passive device according to claim 1, wherein the step of obtaining the geometric parameters of the passive device comprises: receiving user input of the geometric parameters; or obtaining a device type of the passive device; 7. The method of claim 1, wherein, initializing the geometric parameters according to the device type and the performance optimization target by using the optimization algorithm.

9. The automatic design method of the passive device according to claim 8, wherein the step of initializing the geometric parameters comprises: ​ ​ ​ ​ ​ ​ ​ The geometric parameters are generated randomly in a preset space sample; or The fixed geometric parameters are generated according to preset geometric parameters.

10. The method of claim 1, wherein, The optimization algorithm includes a genetic algorithm or a simulated annealing algorithm.

11. The passive device automatic design method of claim 1, wherein The step of optimizing the geometric parameters includes: The optimization algorithm is used to determine an optimization direction of the geometric parameters based on a deviation between the electromagnetic performance parameters and the performance optimization target.

12. The method of claim 1, wherein, The stop condition includes: The number of iterations reaches a preset number; or In a plurality of continuous iterations, a variation of the geometric parameters is less than a preset threshold, or a deviation between the electromagnetic performance parameters and the performance optimization target does not converge.

13. The method of claim 1, wherein, The electromagnetic performance parameters include network parameters or noise coefficients.

14. The method of claim 13, wherein the step of automatically designing the passive device is performed by a computer program. The network parameters include S parameters, Z parameters and / or Y parameters.

15. The method of claim 13, wherein the step of automatically designing the passive device is performed by a computer program. The performance parameters include inductance and / or quality factors.

16. The method of claim 1, wherein, The passive device is an inductor, the geometric parameters include at least one of an inner diameter, a line width, a spacing or a number of turns of the inductor, and the performance optimization target includes at least one of inductance, a quality factor or a noise coefficient.

17. An electronic device, comprising: The electronic device includes: A memory for storing computer executable instructions or computer programs; A processor for executing the computer executable instructions or computer programs stored in the memory to implement the passive device automatic design method of any one of claims 1 to 16.

18. A computer-readable storage medium storing computer-executable instructions or a computer program, wherein the computer-executable instructions or the computer program comprise the steps of: The computer executable instructions or computer programs are executed by the processor to implement the passive device automatic design method of any one of claims 1 to 16. ​ 19. A computer program product comprising computer programs or computer executable instructions, characterized in that, The computer executable instructions or computer programs are executed by the processor to implement the passive device automatic design method of any one of claims 1 to 16. The computer executable instructions or computer programs are executed by the processor to implement the passive device automatic design method of any one of claims 1 to 16.

Images