Topology Optimization Using Approximate FEA for Structural Design
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional topology optimization methods using Finite Element Analysis (FEA) incur high iterative costs due to the need for exact inversions of large system matrices, making them computationally expensive and time-consuming, especially when designing mechanical parts that require fine-grained material distribution and structural optimization.
Innovation Solution
The approach reformulates the problem as a bilevel optimization using a first-order algorithm and the Solid Isotropic Material with Penalization (SIMP) model, allowing for approximate solutions and reducing iterative costs, enabling faster design updates and convergence to locally optimal structures.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional FEA approaches are used for topology optimization, then accurate structural analysis is achieved, but computational time and iterative cost increase significantly
Solution Approach 1:
The patent applies partial action by computing only the necessary components of the FEA solution rather than solving the complete system. Specifically, it calculates only the displacement field and stress components required for sensitivity analysis, omitting redundant computations. This partial computation approach maintains sufficient accuracy for optimization while dramatically reducing computational time and iterative cost.
2Measurement precision
If exact inversion of large FEA system matrices is performed, then precise deformation results are obtained, but algorithmic complexity and computational resources increase
Solution Approach 1:
The patent segments the computational task by dividing the FEA solution process into distinct stages: (1) computing the displacement field, (2) calculating stress components, and (3) performing sensitivity analysis. Each stage processes only the specific data needed for that purpose, avoiding the need to invert the entire large system matrix. This segmentation reduces algorithmic complexity while maintaining precision where required.
Solution Approach 2:
The patent extracts and computes only the essential deformation and stress components needed for topology optimization sensitivity analysis, rather than solving for all possible mechanical parameters. By taking out only the necessary computational elements (displacement and stress fields), the method achieves precise results with reduced algorithmic complexity.
Data Source
AI summary
The various embodiments described herein include methods, devices, and systems for optimizing structural design. In some embodiments, a method includes obtaining a set of constraints for a structure, including an external force constraint; and partitioning a design space for the structure into a plurality of cells. The method further includes, in accordance with the external force constraint being applied to the structure: obtaining an approximate finite element analysis (FEA) solution for the plurality of cells based on the set of constraints; performing a sensitivity analysis on the plurality of cells based on the approximate FEA solution; and updating a structural model for the structure based on the sensitivity analysis of the plurality of cells.