Since the optimal design of dividing wall columns (DWC) is a highly nonlinear and multivariable problem, an appropriate solving tool is required. In this paper a multi-objective genetic algorithm with restrictions is considered to design columns with dividing walls. Also, a methodology is proposed for sizing the DWC. The proposed design methodology allows achieving appropriate designs for columns with two dividing walls. As expected, the physical structures that allow the use of one or two dividing walls are not so different from each other and, as a consequence, the difference in the total annual costs for both systems depends mainly on the energy requirements.