Robust topology optimization accounting for geometric imperfections

Topology optimization is a powerful method to optimize the performance of macro, micro, or nano structures. However, the geometry of the actual structure may differ from the optimized design due to manufacturing errors. Such geometric imperfections can have a significant impact on the structural performance. As a consequence, the actual structure may be far from optimal. In this paper, a robust approach to topology optimization is presented, taking into account two types of geometric imperfections: variations of (1) the crosssections and (2) the locations of structural elements. The first type is modeled by means of a scalar non-Gaussian random field, which is represented as a translation process. The underlying Gaussian field is simulated by means of the EOLE method. The second type of imperfections is modeled as a Gaussian vector-valued random field, which is simulated directly by means of the EOLE method. In each iteration of the optimization process, the relevant statistics of the structural response are evaluated by means of a Monte Carlo simulation. The proposed methodology is successfully applied to a test problem involving the design of a compliant mechanism (for the first type of imperfections) and a vertical load carrying system (for the second type).