Decomposition method for hierarchical Multidisciplinary Robust Design considering uncertainty functions and coupling variables

Exiting decomposition methods based on Function Dependency Table (FDT), those ignore the substantive impact to optimization process brought by performing uncertainty analysis, and cannot decomposes coupling variables, are not suitable for complex Multidisciplinary Robust Design (MRD).To obtain optimal decomposition form for a MRD problem, a novel decomposition method, which can deal with uncertainty functions and coupling variables in MRD, is developed. Compared with existing decomposition methods, this method is more suitable for MRD with two advantages. First, uncertainty functions can be decomposed averagely to obtain better concurrency and less total computational cost. Second, coupling variables can be identified and decomposed appropriately. This method is verified by a gear reducer box problem. Compared with two existing decomposition solutions, this method significantly reduces the total computational cost.