A multi-resolution intelligent optimization method for high-dimensional CAE simulation

By employing a multi-resolution intelligent optimization method, the optimal region is generated and the kernel function is updated in a nested manner. By combining hierarchical orthogonal sampling and hybrid kernel functions, the problems of insufficient computational complexity and model accuracy in high-dimensional complex engineering systems are solved, and efficient and accurate optimization search is achieved.

CN122113595APending Publication Date: 2026-05-29PERA
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
PERA
Filing Date
2026-02-02
Publication Date
2026-05-29

AI Technical Summary

Technical Problem

Existing surrogate optimization methods suffer from high computational costs and insufficient model accuracy when dealing with complex engineering systems with high dimensions and multiple scales. They struggle to find globally optimal hybrid variable design schemes within acceptable computational costs, especially in automotive lightweighting and composite material structure design.

Method used

Employing a multi-resolution intelligent optimization method, this approach generates optimal regions and updates kernel functions through nested processing. By combining hierarchical orthogonal sampling and hybrid kernel functions, it constructs an efficient active learning framework to achieve intelligent progressive focusing in high-dimensional design spaces. It also parallelizes the exploration-development trade-off strategy to handle the collaborative optimization of continuous and discrete variables.

Benefits of technology

It significantly improves the search efficiency and accuracy of high-dimensional optimization problems, reduces the overall optimization cost, ensures the consistency and reliability of the understanding, effectively captures the complex relationships between mixed variables, and avoids local optima traps and wasted computational resources.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122113595A_ABST
    Figure CN122113595A_ABST
Patent Text Reader

Abstract

The application relates to a multi-resolution intelligent optimization method for high-dimensional CAE simulation, and belongs to the technical field of CAE simulation optimization. The method solves the problems of high calculation complexity and single-scale modeling in existing simulation optimization, which leads to slow optimization speed and insufficient prediction accuracy. The method comprises the following steps: determining design variables and their types based on simulation targets of a product to be optimized, and then constructing an initial sample set and an initial surrogate model; performing iterative optimization based on the initial surrogate model until a convergence condition is met; each iteration comprises the following steps: generating a current optimal region and a current local kernel function according to the current surrogate model and the current sample set; updating the kernel function by weighting and superimposing the current local kernel function; sampling new sample points in the current optimal region and obtaining their response values to update the sample set; reconstructing the current surrogate model based on the updated sample set and the updated kernel function; and obtaining a final optimization scheme according to the latest sample set. The method improves the optimization speed and accuracy.
Need to check novelty before this filing date? Find Prior Art