turkmath.org

Products and processes typically have several design objectives that need to be optimized for satisfaction of customers. Designers try to find the most appropriate process design parameters so that products/processes perform as targeted consistently under various different conditions, for customer satisfaction. However, the performance of a product/process changes randomly due to uncontrollable sources of variation. Furthermore, design objectives may depend on each other both statistically and preferentially. These increase complexity of the design optimization problems. Many solution approaches currently available in literature and used in industry convert the multi-response optimization problem into a single objective one using aggregation functions such as quality loss and desirability functions. However, use of these functions poses further difficulties due to their nonlinearity or nonsmoothness in finding their optimal points. In this talk, we discuss and demonstrate drawbacks of these functions and suggest alternative formulations of the problems for optimization of them. This is a joint work with many researchers including G.W. Weber, B. Akteke-Öztürk, L. E. Dolgun and L. Erişkin.