Multifidelity Topology Optimization with Runtime Verification and Acceptance Control: Benchmark Study in 2D and 3D

Area of Science:

• Engineering
• Computational Science

Background:

• Density-based topology optimization demands high-resolution meshes for accuracy, leading to high computational costs, especially in 3D.
• Coarse-mesh methods offer speed but risk discretization errors, potentially leading to suboptimal designs.

Purpose of the Study:

• To develop an efficient and robust multifidelity framework for topology optimization.
• To manage optimizer states and ensure reliable design verification using varying mesh resolutions.

Main Methods:

• A multifidelity framework employing acceptance control for runtime verification.
• Utilizing coarse discretizations for generating design proposals and fine discretizations for verification.
• Implementing a best-referenced criterion for proposal acceptance/rejection and reverting to the last verified state upon failure.

Main Results:

• The framework balances the efficiency of coarse meshes with the accuracy of fine meshes.
• Configurable verification schedules and a cleanup phase optimize performance.
• Evaluated on 2D and 3D benchmark problems, demonstrating effectiveness in compliance, runtime, and design robustness.

Conclusions:

• The multifidelity framework with acceptance control effectively addresses the trade-off between computational cost and accuracy in topology optimization.
• This approach enables reliable and efficient generation of optimized structural topologies, even for complex 3D problems.