Doe based robust optimization considering tolerance bands of design parameters

Due to developments of computer technologies, optimization methods have been issued in areas of computer aided engineering design. These methods provide insight into the actual design process, cost reduction and quality improvement. But, it has been pointed out that general optimization theories based deterministic manners include some problems. For example, deterministic optimization methods do not consider tolerance bands of design parameters (i.e. design variables and material properties). Considering tolerance bands of design parameters, optimization results by deterministic methods might be meaningless. For this reason, a robust design has been researched. The robust design is defined that a product or process is insensitive to the effects of sources of variability, even though the sources themselves have not been eliminated [1]. Generally, there are two solutions to avoid meaningless designs. The one is to minimize the tolerance bands of parameters. But this solution demands a high accuracy and increases production cost. The other one is to calculate the robust optimum with the post-processing deterministic optimization solution. This study presents the robust optimization method considering tolerances of design variables and material properties. To estimate the robustness, this study used analysis of means (ANOM) based orthogonal arrays which is a kind of design of experiments (DOE). And the paper examined the proposed method through mathematical problems and cantilever beam problem. Finally, the paper applied the presented method to a real complex structure.