Optimization for Robust Design

Hosted by ASAP

Professor Peter J Fleming,University of Sheffield

Professor of Industrial Systems and Control.

Abstract

Many real-world design problems can be usefully cast asmulti-criteria search and optimization problems that, in turn, can oftenbenefit from the application of evolutionary computing methods. Thesereal-world problems arise in a variety of forms, such as design exercises,trade studies or allocation problems. The sought outcome is a single solution,often a compromise that takes account of conflicting criteria and is robust tounderlying assumptions.

Such a problem-solving process has at its core a search andoptimization algorithm. However, first, problem requirements must be capturedand appropriate models identified to enable the evaluation of objectives andconstraints. A search algorithm is applied to generate potential solutions andthis stage is followed by an interactive data analysis decision-making stageleading to the problem solution.

Each of these phases generates its own research challenges.Using real-world problem examples, these challenges will be discussed, somesolutions suggested and other questions left open, such as … what if there area number of these designs being undertaken in parallel that have dependencieson one another?

Peter Fleming is Professor of Industrial Systems and Control in the Department of Automatic Control and Systems Engineering.  His systems and control engineering research interests are motivated by the use of multi-criteria decision-making approaches in problem-solving.  Research includes algorithm development, decision-making strategies and the management of exogenous and endogenous uncertainty in design optimization.  Further details may be found at http://www.shef.ac.uk/acse/staff/peter_fleming