A global two-layer meta-model for response statistics in robust design optimization

Robust design optimization (RDO) of large-scale engineering systems is computationally intensive and requires significant CPU time. Considerable computational effort still required within conventional meta-model assisted RDO frameworks. The primary objective this article to minimize further the requirements by developing a global two-layered approximation based technique. in inner layer approximates response quantity outer statistics computed from meta-model. This approach eliminates both model building Monte Carlo simulation cycle, considerably fewer actual evaluations than single-layered approximation. To demonstrate approach, two recently developed compressive sensing enabled globally refined Kriging models have been utilized. proposed framework applied one test example real-life applications illustrate clearly its potential yield robust optimal solutions with minimal cost.