Decomposed multilevel optimization and bayesian uncertainty quantification of nano-enhanced composite sandwich plates
Through a combination of Bayesian uncertainty quantification and decomposed multilevel optimization, a non-deterministic approach is presented for application to material-product design problems. Consistent with the integrated computational materials engineering philosophy, the hierarchical decomposition and solution strategy combines micro- and macrolevel material modeling and design with structural level analysis and optimization for a honeycomb sandwich plate with carbon nanofiber-enhanced, fiber-reinforced polymer composite facesheets. Particular emphasis is placed on the Bayesian modeling of epistemic uncertainty in the orientation distribution of the nano-inclusions at the micro-level based on a belief structure with multiple focal elements. Several failure modes are considered, including first-ply failure in the macro-material level, global buckling, shear crimping, wrinkling, and dimpling failure in the structural-level analysis. The composite sandwich plate provides an indepth demonstration of how epistemic uncertainty modeled using Bayesian updating and aleatory uncertainties may be combined for non-deterministic, decomposed engineering design optimization.
Citation: Dettwiller-, I., and Rais-Rohani, M., “Decomposed multilevel optimization and bayesian uncertainty quantification of nano-enhanced composite sandwich plates,” AIAA SciTech 2017: 19th AIAA Non-Deterministic Approaches Conference, Grapevine, TX, Jan 9-13, 2017.