A decomposition based design method coordinated by disciplinary subspace optimization

The paper describes the development of a decomposition based design (DBD) method using optimal sensitivity information with respect to coupled or interdisciplinary design variables. A multidisciplinary design system considered in the present study is decomposed into a number of subspaces based on their own design objectives and constraints associated with engineering physics. The coupled relations among subspaces are identified by interdisciplinary design variables. Subsystem level optimization is performed in parallel, and the system level coordination is determined by the first order optimal sensitivities of subsystem level objective functions with respect to interdisciplinary design variables. The central of the present work focuses on the formulation of system level coordination strategy and its capability in decomposition based design. A fluid-structure coupled design problem is explored as a test-bed to verify the proposed DBD method.