In the present paper, comparisons of solving a cyclic chromatographic separation problem using Mixed Integer Nonlinear Programming (MINLP) methods are presented. The dynamics of the chromatographic separation process is modeled as a boundary value problem that is solved, repeatedly within the optimization, using a relatively fast and numerically robust finite difference method. The MINLP methods considered are the Extended Cutting Plane (ECP), the Branch and Bound (BB), and the Outer Approximation (OA) methods. The comparisons indicate the advantages of the ECP method that needs relatively few function evaluations. The results also show that the total efficiency of traditional OA and BB methods, that solves a sequence of Nonlinear Programming (NLP) subproblems, is substantially degradated in these kinds of optimization problems involving both a combinatorial task and time-consuming numerical evaluations. (C) 2004 Elsevier Ltd. All rights reserved.