文摘
Identifying the relevant sources of uncertainty is a relevant issue in different problems, including topology optimization, as too many sources of uncertainty investigated at too many levels can make the computational effort prohibitive. Hence, it is fundamental to develop schemes to identify those parameters and parameter interactions which cause the most change in the final problem’s results. In this article, a simple hierarchical procedure is proposed for parameter identification and it is illustrated in a robust topology optimization example. The scheme is composed of a pre-screening of candidate parameters through a component of variation analysis. Only the most relevant parameters, identified in the initial screening, are forwarded to a design of experiments (DOE) analysis. In the DOE, a subset of most relevant parameters and parameters’ interactions are identified, and forwarded to the last stage: the robust optimization. Having identified the most relevant contributions, in the robust optimization one can afford to increase the number of levels for the most relevant parameters, hence also capturing non-linear effects. The example application problem is presented to illustrate the proposed scheme and to investigate the potential gain in terms of computational effort reduction.