CN-121997722-A - Multi-objective optimization method and device for electromagnetic transmission characteristics of coaxial through silicon via array

CN121997722ACN 121997722 ACN121997722 ACN 121997722ACN-121997722-A

Abstract

The invention discloses a multi-objective optimization method of electromagnetic transmission characteristics of a Coaxial Through Silicon Via (CTSV) array, which comprises the steps of constructing a full-wave simulation model of the CTSV array based on physical structures, material parameters, geometric structure parameter combinations and scattering parameters of the CTSV array to be optimized, constructing a data set based on the simulation model, constructing a GRNN network model framework by utilizing the data set, determining an optimal smoothing factor of the model framework by utilizing a particle swarm optimization algorithm, constructing and training a GRNN model based on the smoothing factor and the data set to obtain a trained GRNN model, iteratively solving the geometric structure parameter combinations of the CTSV array based on a genetic algorithm and the trained GRNN network model, and outputting updated geometric structure parameter combinations when preset conditions are met. The invention can improve the accuracy of designing the CTSV array parameters, can also reduce the calculation cost and improve the efficiency of optimizing the CTSV array parameters.

Inventors

DONG GANG
HONG YUHENG
Zhi Changle
ZHU ZHANGMING

Assignees

西安电子科技大学

Dates

Publication Date: 20260508
Application Date: 20260105

Claims (10)

1. The multi-objective optimization method for the electromagnetic transmission characteristics of the coaxial through silicon via array is characterized by comprising the following steps of: based on the physical structure, material parameters, optimized geometric structure parameter combination and scattering parameters of the coaxial through silicon via array to be optimized, constructing a full-wave simulation model of the coaxial through silicon via array; Sampling the optimized geometric structure parameter combination based on the full-wave simulation model to obtain a plurality of groups of geometric structure parameter combinations and electrical performance indexes corresponding to each group of geometric structure parameter combinations, and constructing the plurality of groups of geometric structure parameter combinations and the electrical performance indexes corresponding to each group of geometric structure parameter combinations into a data set; Taking a geometric structure parameter combination in the data set as input, taking a corresponding electrical performance index as output, constructing a generalized regression neural network model framework, determining an optimal smoothing factor of the generalized regression neural network model framework by using a particle swarm optimization algorithm, and constructing and training a generalized regression neural network model based on the optimal smoothing factor and the data set to obtain the trained generalized regression neural network model; And carrying out iterative solution on the geometric structure parameter combination of the coaxial through silicon via array based on a genetic algorithm and the trained generalized regression neural network model, and outputting the updated geometric structure parameter combination to be used as the design parameter of the coaxial through silicon via array when the preset iteration termination condition is met.
2. The method for multi-objective optimization of electromagnetic transmission characteristics of a coaxial through-silicon via array according to claim 1, wherein the constructing a full-wave simulation model of the coaxial through-silicon via array based on physical structure, material parameters, optimizable combination of geometric parameters, and scattering parameters of the coaxial through-silicon via array to be optimized comprises: Based on the physical structure, material parameters and optimized geometric structure parameter combination of the coaxial through silicon via array, establishing an electromagnetic simulation model of the three-dimensional integrated circuit; And calibrating the electromagnetic simulation model based on the scattering parameters to obtain a full-wave simulation model of the coaxial through silicon via array.
3. The method for optimizing electromagnetic transmission characteristics of a coaxial through-silicon-via array according to claim 1, wherein the sampling the optimizable combination of geometry parameters based on the full-wave simulation model to obtain a plurality of groups of combinations of geometry parameters and electrical performance indexes corresponding to each group of combinations of geometry parameters comprises: sampling the geometric structure parameter combination based on the full-wave simulation model to obtain a plurality of groups of sample parameter combinations; And performing full-wave simulation on each group of sample parameter combinations to obtain electrical performance indexes corresponding to each group of geometric structure parameter combinations.
4. The multi-objective optimization method of electromagnetic transmission characteristics of a coaxial through-silicon via array according to claim 1, wherein the data set is divided into a training set and a verification set according to a preset proportion; the method for determining the optimal smoothing factor of the generalized regression neural network model framework by using a particle swarm optimization algorithm, and constructing and training a generalized regression neural network model based on the optimal smoothing factor and the data set to obtain the trained generalized regression neural network model comprises the following steps: Initializing a particle swarm, setting a search range of smoothing factors of the initial generalized regression neural network model framework, and randomly initializing the position and the speed of each particle in the particle swarm, wherein the position of each particle represents a candidate smoothing factor; performing iterative optimization on the particle swarm, and performing the following steps in each iteration: S3.1, obtaining smoothing factors represented by the current position of each particle in the particle swarm, constructing and training based on the smoothing factors corresponding to each particle and the training set to obtain a corresponding generalized regression neural network model, inputting the geometric structure parameter combination in the verification set into the generalized regression neural network model corresponding to each particle to obtain an electrical performance index predicted by each generalized regression neural network model, calculating an error between the electrical performance index predicted by each generalized regression neural network model and a real electrical performance index corresponding to the verification set, and taking the error as a current fitness value of the corresponding particle; s3.2, determining a global optimal position and an individual optimal position of each particle based on the current fitness value, the historical optimal fitness value and the historical global optimal position of each particle, wherein the global optimal position is a position corresponding to the minimum fitness value of the particle determined by the particle swarm in the iterative process; S3.3, updating the current speed and the current position of each particle based on the individual optimal position of each particle, the global optimal position, the speed update formula of the particle swarm optimization algorithm and the position update formula; And S3.4, repeatedly executing S3.1-S3.3 until the preset maximum iteration times are reached, taking a smoothing factor corresponding to the current global optimal position as the optimal smoothing factor, and constructing and training a generalized regression neural network model based on the optimal smoothing factor and the training set to obtain the trained generalized regression neural network model.
5. The method of claim 4, wherein determining the global optimal position and the individual optimal position for each particle based on the current fitness value, the historical optimal fitness value, and the historical global optimal position for each particle comprises: Comparing the current fitness value of the particle with the historical optimal fitness value for each particle, if the current fitness value is better, determining the current position as the individual optimal position of the particle, otherwise, keeping the original individual optimal position; And determining a target individual optimal position in the individual optimal positions of all particles based on the individual optimal positions of all particles and the corresponding fitness values, comparing the target individual optimal position with a historical global optimal position, determining the target individual optimal position as the global optimal position if the target individual optimal position is better, and otherwise, keeping the historical global optimal position.
6. The method for optimizing electromagnetic transmission characteristics of a coaxial through-silicon via array according to claim 1, wherein the iterative solution is performed on the geometric parameter combination of the coaxial through-silicon via array based on a genetic algorithm and the trained generalized regression neural network model, and when a preset iteration termination condition is satisfied, the updated geometric parameter combination is output and used as a design parameter of the coaxial through-silicon via array, and the method comprises: Encoding the geometric parameter combinations by using the genetic algorithm, and taking the geometric parameter combinations as chromosome representations to construct an initial population, wherein each individual in the initial population corresponds to a group of geometric parameter combinations; predicting electrical performance indexes corresponding to each individual in the initial population by using the trained generalized regression neural network model; constructing a multi-objective fitness function based on the predicted electrical performance index, the multi-objective fitness function being used to characterize fitness of each individual; And executing genetic operation in the genetic algorithm to iteratively update the geometric structure parameter combination in the initial population, and outputting the updated geometric structure parameter combination as the design parameter of the coaxial through-silicon-via array when a preset iteration termination condition is met.
7. The method of claim 6, wherein the genetic operations include a selection operation based on roulette strategy, a crossover operation based on double-point crossover, and a mutation operation based on random bit flipping.
8. The method of claim 6, wherein the electrical performance metrics include return loss, insertion loss, near-end crosstalk, and far-end crosstalk; The multi-objective optimization function is a function obtained by carrying out weighted summation on the return loss, the insertion loss, the average value of the near-end crosstalk and the average value of the far-end crosstalk.
9. The method of claim 5, wherein the iteration termination condition comprises reaching a predetermined maximum number of iterations or a magnitude of change in population fitness less than a predetermined threshold.
10. A multi-objective optimization apparatus for electromagnetic transmission characteristics of a coaxial through-silicon via array, comprising: The construction module is used for constructing a full-wave simulation model of the coaxial through silicon via array based on the physical structure, the material parameters, the optimized geometric structure parameter combination and the scattering parameters of the coaxial through silicon via array to be optimized; The construction module is further used for sampling the optimized geometric structure parameter combination based on the full-wave simulation model to obtain a plurality of groups of geometric structure parameter combinations and electrical performance indexes corresponding to the geometric structure parameter combinations, and constructing the geometric structure parameter combinations and the electrical performance indexes corresponding to the geometric structure parameter combinations as a data set; The model training module is used for taking the geometric structure parameter combination in the data set as input and the corresponding electrical performance index as output, constructing a generalized regression neural network model framework, determining the optimal smoothing factor of the generalized regression neural network model framework by utilizing a particle swarm optimization algorithm, and constructing and training a generalized regression neural network model based on the optimal smoothing factor and the data set to obtain the trained generalized regression neural network model; And the optimization design module is used for carrying out iterative solution on the geometric structure parameter combination of the coaxial through silicon via array based on a genetic algorithm and the trained generalized regression neural network model, and outputting the updated geometric structure parameter combination to be used as the design parameter of the coaxial through silicon via array when the preset iteration termination condition is met.

Description

Multi-objective optimization method and device for electromagnetic transmission characteristics of coaxial through silicon via array Technical Field The invention belongs to the technical field of three-dimensional integrated circuits, and particularly relates to a multi-objective optimization method and device for electromagnetic transmission characteristics of a coaxial through silicon via array. Background With the rapid development of Three-dimensional integrated circuit (Three-Dimensional Integrated Circuit,3D IC) technology, signal transmission in the terahertz frequency band puts higher performance demands on Through-Silicon Via (TSV) interconnection structures. A Coaxial Through-Silicon Via (CTSV) array exhibits significant advantages in high frequency, high speed interconnects due to its excellent shielding characteristics. However, the electromagnetic transmission performance of CTSV arrays is strongly affected by their geometric parameters (e.g., aperture, pitch, dielectric layer thickness, etc.), and there are complex interrelationships between multiple performance metrics (e.g., return loss, insertion loss, near-end crosstalk, and far-end crosstalk). This results in the design of CTSV arrays in the terahertz frequency band as a typical multi-objective, strongly nonlinear, high-dimensional optimization problem. Currently, CTSV array design mainly relies on full-wave electromagnetic simulation tools (e.g. ANSYS HFSS) based on finite element method, and performance evaluation and optimization is performed by parameter scanning or trial and error method. Although full-wave simulation can provide higher precision, the calculation cost is huge, the simulation period is long, and particularly when multi-parameter combination and multi-objective optimization are processed, massive simulation iteration is needed, so that the design efficiency is low and the time cost is high. In addition, the traditional equivalent circuit model or analysis method is often insufficient in precision in a high-frequency complex electromagnetic coupling scene, so that the mapping relation between parameters and performance is difficult to accurately describe, and the applicability of the equivalent circuit model or analysis method in the optimization design is limited. Therefore, how to improve the accuracy of the mapping relationship between the design parameters and the performance and reduce the calculation cost when designing the parameters of the CTSV array, thereby improving the efficiency of optimizing the parameters of the CTSV array is a technical problem to be solved at present. Disclosure of Invention Aiming at the problem of improving the efficiency of optimizing the parameters of the CTSV array by improving the accuracy of the mapping relation between the design parameters and the performance and reducing the calculation cost at the same time when the parameters of the CTSV array are designed, the invention provides a multi-objective optimization method and device for the electromagnetic transmission characteristics of the coaxial through silicon via array. The technical problems to be solved by the invention are realized by the following technical scheme: the invention provides a multi-objective optimization method for electromagnetic transmission characteristics of a coaxial through silicon via array, which comprises the following steps: Based on the physical structure, material parameters, optimized geometric structure parameter combination and scattering parameters of the coaxial through-silicon-via array to be optimized, a full-wave simulation model of the coaxial through-silicon-via array is constructed; Based on a full-wave simulation model, sampling the optimized geometric structure parameter combinations to obtain a plurality of groups of geometric structure parameter combinations and electrical performance indexes corresponding to each group of geometric structure parameter combinations, and constructing the plurality of groups of geometric structure parameter combinations and the electrical performance indexes corresponding to each group of geometric structure parameter combinations into a data set; The method comprises the steps of taking a geometric structure parameter combination in a data set as input, taking a corresponding electrical performance index as output, constructing a generalized regression neural network model framework, determining an optimal smoothing factor of the generalized regression neural network model framework by using a particle swarm optimization algorithm, and constructing and training a generalized regression neural network model based on the optimal smoothing factor and the data set to obtain a trained generalized regression neural network model; And (3) carrying out iterative solution on the geometric structure parameter combination of the coaxial through silicon via array based on a genetic algorithm and a trained generalized regression neural network model, and outp