Optimization and Systems Theory Seminar

June 1, at 11.00, room 3721, Lindstedtsvägen 25, KTH:
Jussi Hakanen, Department of Mathematical Information Technology, University of Jyväskylä, Finland

Interactive Multiobjective Optimization - Applications in Chemical Process Industry

Practical optimization problems typically involve several criteria that are conflicting (e.g. cost, quality and environmental effects) and thus can not be optimized simultaneously. Interactive multiobjective optimization methods have succesfully been used in solving many such practical applications where the aim is to find a most preferred compromise between the conflicting criteria. The main idea is to search for a best compromise solution through an iterative solution process where the roles of a decision maker, DM, (or a designer) providing preference information and the optimization system alternate. This interaction enables both the DM learning about the interdependencies between the conflicting evaluation criteria and utilizing the experiences of the DM in guiding the search towards interesting and preferred regions. In addition, with this approach we can avoid computing such compromises that are not interesting to the DM, which can lead to significant savings on computing time in the cases of computationally costly problems. In this presentation, applications of interactive multiobjective optimization in the field of chemical process industry are described including their practical challenges. Applications are related to paper making, chemical separation processes and wastewater treatment plant design. These problems have been solved with the IND-NIMBUS multiobjective optimization framework developed at the University of Jyväskylä. IND-NIMBUS can be connected to different modelling and simulation tools describing the objectives of the phenomena to be optimized. From the optimization point of view, IND-NIMBUS contains different types of interactive multiobjective optimization methods that can be used for problems having different challenges (e.g. computational cost vs. accuracy, non- convexity, how the DM wants to express preferences). In addition, IND-NIMBUS has tools for visualizing and storing the compromise solutions computed.

Calendar of seminars Last update: May 29, 2012.