Most optimization-based approaches in conceptual, process design either focus on global optimization using simplified process models or utilize some kind of metaheuristic to optimize by means of repetitive runs of a detailed simulation model. Because the design of even a single distillation column model results in a nonconvex large-scale and mixed-integer optimization problem, if rigorous thermodynamic models are applied, deterministic optimization is still mostly limited to local optimization. In order to investigate and improve the solution quality of previously developed efficient local optimization approaches, this paper proposes a hybrid evolutionary deterministic optimization approach. The resulting memetic algorithm not only allows the optimization of the initial process structure but also facilitates discrete decision making that severely complicates a deterministic optimization due to the resulting discontinuities. The proposed approach not only eases the application by reducing the necessary user input for the initialization but also strengthens the confidence in the quality of the results because it provides an extensive screening of the design space. Several case studies, including utility and entrainer selection, demonstrate the performance of the hybrid optimization approach and suggest that even more complex design problems can be solved efficiently.