Optimization and Systems Theory Seminar
Thursday, September 27, 2001, 13.00-14.00, Room 3721, Lindstedtsvägen 25

Sven Leyffer
Department of Mathematics
University of Dundee
Dundee, Scotland, UK
E-mail: sleyffer@maths.dundee.ac.uk

A review of mixed integer nonlinear programming

Many optimization problems involve variables which are restricted to take binary, integral or discrete values. Applications include chemical engineering such as process synthesis, batch plant design, cyclic scheduling or the design of distillation columns. More recently, applications have arisen in the nuclear industry, to optimize core reload operations and in topology optimization, where binary variables model the presence or absence of material in each finite element.

This talk surveys the recent developments in the design of solvers for large Mixed Integer Nonlinear Programming (MINLP) problems. It will start by reviewing classical methods such as branch-and-bound, Benders Decomposition and Outer Approximation. Next, new hybrid approaches, combining these classical methods are discussed and likely future developments are pointed out.

Finally, the solution of MINLPs by a parallel implementation of branch-and-bound is considered. An important feature of this implementation is the use of a computational grid or meta-computer as the underlying computing platform. Computational experience on a set of large MINLPs is reported which indicates that this approach is efficient for the solution of large MINLPs.

Calendar of seminars
Last update: September 14, 2001 by Anders Forsgren, anders.forsgren@math.kth.se.