Robust design optimization of nanostructure for solar thermal absorption

Abstract

Recently, nanostructure is being introduced to solar thermal absorber for its outstanding performance. However, nanostructure has a fabrication uncertainty causing inaccurate design and thus unexpected performance. In this study, robust design optimization (RDO) of nanostructure for solar thermal absorber is sought to resolve the uncertainty issue with two new RDO methodologies: Hessian-eigenvector dimension reduction (HeDR) method and Basis screening (BS) Kriging. Dimension reduction (DR) method is widely used for estimating statistical moments of a performance model with N-dimensional random variables as the input parameters for its computational efficiency. DR method reduces an N-dimensional integration into N decomposed 1-D numerical integrations. The decomposition deletes the effect of cross-terms and causes a falling off in the computation accuracy. However, the error can be minimized with HeDR method setting sample point pattern transformed into the directions the eigenvector of Hessian matrix indicate. Metamodeling is widely used for simulation-based design optimization for computationally heavy engineering models. The dynamic Kriging method is one of the most popular metamodeling methods due to its accuracy. BS Kriging method, an improved dynamic Kriging method seeks a promising basis function set in a more efficient way with 3 additional technical strategies: 1. Limitation of the maximum order of basis functions, 2. Estimation of the importance of basis terms, 3. Determination of the basis function using cross-validation error. Both methods will be numerically verified and utilized to obtain the robust optimal design of nanostructure for solar thermal absorption. RDO of 1-D and 2-D pattern nanostructure will be conducted with HeDR method and BS Kriging method, respectively.