A cluster mean approach for topology optimization of natural frequencies and bandgaps with simple/multiple eigenfrequencies

TL;DR

This paper introduces a cluster mean approach to ensure differentiability in topology optimization problems involving multiple eigenfrequencies and bandgaps, enabling smooth optimization of complex structures.

Contribution

It proposes a novel cluster mean method that guarantees differentiability of multiple eigenvalues, improving the reliability of eigenfrequency topology optimization.

Findings

Method achieves smooth convergence in 2D and 3D examples.

Cluster mean approach handles simple and multiple eigenvalues effectively.

Optimization of eigenfrequencies and bandgaps is successfully demonstrated.

Abstract

This study presents a novel approach utilizing cluster means to address the non-differentiability issue arising from multiple eigenvalues in eigenfrequency and bandgap optimization. By constructing symmetric functions of repeated eigenvalues -- including cluster mean, p-norm and KS functions -- the study confirms their differentiability when all repeated eigenvalues are included, i.e., clusters are complete. Numerical sensitivity analyses indicate that, under some symmetry conditions, multiple eigenvalues may also be differentiable w.r.t the symmetric design variables. Notably, regardless of enforced symmetry, the cluster mean approach guarantees differentiability of multiple eigenvalues, offering a reliable solution strategy in eigenfrequency topology optimization. Optimization schemes are proposed to maximize eigenfrequencies and bandgaps by integrating cluster means with the bound formulations. The efficacy of the proposed method is demonstrated through numerical examples on 2D and 3D solids and plate structures. All optimization results demonstrate smooth convergence under simple/multiple eigenvalues.