Optimization, Based On The Reliability Of The Design Of Collaborative Optimization And Single Discipline Feasible Method

Design optimization is a crucial element in engineering design. Most modern engineering design problems are related to large scale, complex and multi-disciplinary systems. Therefore, a very active research area called Multidisciplinary Design Optimization (MDO) has been developed specifically for the problem of this kind. The main idea of MDO is to consider the coupling effects among different disciplinary and reach the optimized solution through interactions and collaborations among all areas. However, in engineering practice, uncertainties associated with many important engineering factors, such as applied loadings, material properties, component geometry and other environmental factors, must be considered when we are formulating an optimization model and solving for the optimal solution. Hence, it is very desirable to include uncertainties into MDO for realistic engineering problems.Conventionally, a"safety factor"approach has been used. Although this method can simplify the design process and reduce the computational cost, it sometimes arrives at an over-designed solution that compromised the design performance for design safety. To this end, we must develop a new optimization method that is based on reliability so that the optimized solution will satisfy the required levels of uncertainty.There are a few important research topics in this thesis:(1) Reliability Based Single-disciplinary Design Optimization:Based on the concept of Collaborative Optimization (CO), single disciplinary reliability based design optimization problems with multiple constraints are converted into a series of RBDO problems with single constraint and the original objective function is optimized at the upper level. A cantilever design example is presented to demonstrate the effectiveness of this approach.(2) Reliability analysis under Multi-disciplinary systems:Because Reliability analysis is very expensive during Reliability Based Multi-disciplinary Design Optimization (RBMDO), a method using CO is proposed. Numerical examples are included to illustrate its efficiency. (3) Reliability Based Multi-disciplinary Design Optimization:Due to the complexity of multi-disciplinary systems, when consider reliability in MDO, the problem becomes very complicated. We incorporated Sequential Optimization and Reliability Assessment (SORA) into RBMDO in order to improve its computational efficiency.