Multilevel collaborative aerodynamic design optimization based on Sobol' global sensitivity analysis

Surrogate model combined with global optimization algorithm is necessary for design space exploration in aerodynamic shape optimization (ASO). However, the “curse of dimensionality” exists to a great extent in those global optimization algorithms. Multilevel collaborative optimization (MCO) method is studies to cope with high-dimensional optimization problems in this paper. The superiority of MCO method over traditional direct full-variables optimization method is confirmed through different test functions. In aerodynamic shape optimization, Sobol' global sensitivity analysis is introduced to quantify the importance degrees of design variables. The design variables are divided in to subcomponents according to their importance degrees and the subcomponents are optimized individually in multiple cycles. The MCO aerodynamic design framework is established by integrating the Sobol' global sensitivity analysis method, efficient shape parameterization method, mesh deformation technique, numerical simulation method and surrogate-based global optimizer. Finally, a commercial airplane in transonic regime is optimized by MCO method and conventional method respectively. Results show that the proposed MCO method is better than conventional method.