Model-based decomposition synthesis using Integer Programming for Non-Deterministic MDO

A decomposition synthesis method using Integer Programming for Non-Deterministic Multidisciplinary Design Optimization is developed. Firstly, an Uncertainty Function Dependency Table (UFDT) is advanced by investigating influence that variables' uncertainty brings to decomposition, and the relation between elements of UFDT and uncertain variables is formulated. Secondly, because 'task size' is the focus in decomposition, in order to measure 'task size' more accurately subject to uncertainty, new definition and formulations of the task size are given based on UFDT. Thirdly, Integer Programming algorithm is used to partition initial design problem into a main task and several subtasks. The Integer Programming model is constructed by analyzing decomposition's objectives constrains and variables, and Genetic Algorithm is applied to deal with model's nonlinearity. Finally, a classic MDO test suite is decomposed by applying the developed method, and the results are compared against a traditional method without considering uncertainty. The comparison indicates that developed method can obtain better decomposition.