Multidisciplinary Design Optimization Under Uncertainties

With the development of science and technology, engineering system becomes more and more complex. Design of the system will involve multi-discipline field, and there are coupling relationships among disciplines. Multidisciplinary design optimization (MDO) is a synthesis method for design of complex and coupled system: with the aim of improving performance, reducing cost and shortening design period, it fully utilizes the coupling relationships among disciplines, and proceeds design optimization from system point of view to obtain the global optimum of a complex and coupled system.In practical complex and coupled system, uncertainty widely exists, and it can be divided into aleatory uncertainty and epistemic uncertainty. For achieving high reliable and safe complex and coupled system, the consideration of uncertainty's effect to the system has been a focus of engineering system design. For the case of continuous variables and parameters associated with aleatory uncertainty, Reliability Based Multidisciplinary Design Optimization (RBMDO) has gained increasing attention, and lots of achievements have been achieved. In this paper, based on possibility theory, different design optimization problems are studied:MDO problem with continuous variables and parameters associated with epistemic uncertainty; MDO problem with continuous variables and parameters associated with both aleatory and epistemic uncertainties; in single disciplinary, design optimization problem with discrete and continuous variables and parameters associated with both aleatory and epistemic uncertainties; MDO problem with discrete and continuous variables and parameters associated with both aleatory and epistemic uncertainties. The research in this paper extends and develops the study of MDO under uncertainty, and also provides effective design methods for multi-discipline coupling system.The contributions of this dissertation are summarized as follows:(1) Possibility Based Multidisciplinary Design Optimization (PBMDO). In MDO, when continuous variables and parameters are associated with epistemic uncertainty, the uncertainty analysis method is proposed, and two methods for solving PBMDO problems are developed, named as PBMDO in framework of Sequential Optimization and Reliability Assessment (PBMDO-SORA) and PBMDO in framework of Safety Factor based Approach (PBMDO-SFA), respectively.(2) MDO under both aleatory and epistemic uncertainties (Mixed Variables MDO, MVMDO). In MDO, when continuous variables and parameters are associated with both aleatory and epistemic uncertainties, the uncertainty analysis method is proposed, and also two methods for solving MVMDO problems are developed, named as MVMDO in framework of SORA (MVMDO-SORA) and MVMDO in framework of SFA (MVMDO-SFA), respectively.(3) Random/Fuzzy Continuous Discrete Variables Design Optimization (RFCDV-DO). In single discipline, when continuous and discrete variables and parameters are associated with both aleatory and epistemic uncertainties, based on conditional possibility of failure, two uncertainty analysis method are developed, and a method, named as Random/Fuzzy SORA (RFSORA), is developed to solve RFCDV-DO problems.(4) Random/Fuzzy Continuous Discrete Variables MDO (RFCDV-MDO). In MDO, when continuous and discrete variables and parameters are associated with both aleatory and epistemic uncertainties, based on developed uncertainty analysis methods in single discipline, two uncertainty analysis methods are proposed, and also a method of RFCDV-MDO in framework of SORA (RFCDV-MDO-SORA) is developed to solve RFCDV-MDO problems.