This paper explores an accurate and computationally efficient decomposition method, known as the adaptive sparse polynomial dimensional decomposition (AS-PDD), for uncertainty quantification and design under uncertainty of complex engineering systems. Unlike the truncated polynomial dimensional decomposition requiring their truncation parameter(s) to be assigned apriori or arbitrarily, the AS-PDD performs these truncations automatically by progressively drawing in higher-variate or higher-order contributions as appropriate. Two adaptive sparse schemes are explored in this study. The first scheme arises from the variance-based index. The second scheme utilizes the f -index, which is capable of accounting for the entire probability distribution of the output. Shape design of a three-hole bracket with nine random parameters was performed, demonstrating the power of the new method developed to tackle practical robust design optimization problems.
Available at: http://works.bepress.com/xuchun_ren/19/