Statistical topology optimization for damage identification for orthotropic and cellular structures

Abstract

This study aims to enhance the accuracy and robustness of structural damage identification by extending the statistical topology optimization (STO) framework. While previous STO research has primarily focused on isotropic materials, its applicability to orthotropic and cellular structures has not been fully explored. To broaden its scope, the approach applies the STO framework to models with directional stiffness and periodic microstructures. Multiple topology optimization runs are performed under varied frequency excitations, and consistent damage patterns are extracted using density-based spatial clustering (DBSCAN). Unlike earlier studies, this work introduces genetic algorithm-based tuning of DBSCAN parameters to improve clustering reliability and reduce user dependency. Damage is modeled differently according to the structure type: through density reduction or principal direction rotation in orthotropic models, and by adjusting the void size within cellular unit cells, from which the effective material properties are derived through polynomial-based numerical homogenization. Numerical examples confirm that the framework accurately localizes damage under complex material conditions and achieves superior performance compared to conventional methods.