(534g) Global Optimization of Nonconvex MINLP Problems by Domain and Image Partitioning with Applications to Heat Exchanger Networks

Global Optimization of Nonconvex MINLP Problems by Domain and Image Partitioning with Applications to Heat Exchanger Networks

Sung Young Kim, Debora Faria and *Miguel Bagajewicz

School of Chemical Engineering and Material Science, University of Oklahoma

100 East Boyd Street, T-335 – Norman, OK 73019-0628 USA

* Corresponding Author

Abstract

We propose a new method to obtain the global optimum of full nonconvex MINLP problems. The method is based on partitioning the domain and  image of nonconvex functions. The procedure we propose uses an MILP lower bound constructed using domain/image partitioning. Then a newly developed bound contraction procedure is applied and compared to branch and bound as well as branch and bound  with bound contraction at each node. To illustrate the method we focus on the heat exchanger network stage-wise model. Results show a robust behaviour where the solution does not need initial values as local MINLP model like Dicopt need.  In the presentation we will compare the performance of all options.