An algorithm for performing robust design optimization via use of observational data

Starting from the theory of D and T-optimality criteria, the algorithm links the sequential approach which is typical of a T-optimal design and the combined-array procedure through statistical modeling. The final goal is the achievement of the optimal solution in terms of location and dispersion effects, i.e. by minimizing the dispersion together with the adjustment to the target value, defined as the expected value of the response. The full procedure is illustrated through a simulated numerical case study deriving from an actual industrial engine process.