Date of Graduation
The problem of robust design and optimization for designs with varying operating conditions is considered in this work. A robust design optimization tool (RDOT) has been developed in this work. The tool is constructed with building blocks: an Artificial Neural Network, a Response Evaluator (e.g., FEA, etc.), an Optimizer based on Quadratic Programming, Cooperative Game Theory, Design of Experiments, and Taguchi Methods. The tool is fully codified into a computer program, and it demands very little designer insight into the problem under consideration. Design optimization is necessary to efficiently accomplish a design that, while satisfying the desired functionality, also uses minimum amount of resources. Robust design optimization (RDO) is designing such a product to exhibit insensitivity to the variations in the operating conditions. The objective of RDOT is to help the designer identify the settings of the design variables that result in a robust product. The current optimization methodology has been illustrated with composite material structures to highlight the advantages of RDOT. These structures do not possess a closed form solution for the responses (e.g., deflection, strength, etc.) and changes in the design variables and operating conditions often produce conflicting changes in the responses. The utility of RDOT has been verified with the help of several examples. RDOT also deals with problems whose response surfaces are non-convex. The tool can be used in a generic engineering design problem whose response evaluation is either too expensive to compute analytically/numerically, or can only be evaluated experimentally.
Makkapati, Satheesh, "Robust design optimization with applications to composite-material structures." (1997). Graduate Theses, Dissertations, and Problem Reports. 9348.