An efficient solution method for the MINLP optimization of chemical processes

Process synthesis often involves the solution of large nonlinear discretecontinuous optimization problems, which are usually formulated as mixedinteger nonlinear programming (MINLP) or generalized disjunctive programming (GDP) problems and solved with MINLP solvers. This paper presents an efficient solution method for these problems named successive relaxed MINLP (SR-MINLP), where the model formulations are reformulated to contain only continuous variables. The discrete decisions are relaxed and successively tightened in a sequential solution procedure to facilitate convergence and to obtain local optima of good quality. The solution method is illustrated by a simple numerical example as well as a large and complex example from process synthesis.