Interactive multiobjective optimization of engineering systems
Ravindra V. Tappeta, John E. Renaud · 1999
Many engineering design problems are multiobjective in nature, where designers work to maximize or minimize several objectives simultaneously, while satisfying design constraints. These objectives are potentially conflicting design goals imposed on the technical and economical performances of a given system design. This research focuses on multiobjective system design and optimization. The primary goal of this research is to develop mathematically rigorous and computationally efficient multiobjective optimization strategies which involve the designer in decision making for both single and multidisciplinary systems. A multiobjective, single discipline, systems design strategy called the interactive MultiObjective Optimization Design Strategy (or iMOODS) is developed in this thesis. This strategy provides the Decision Maker (DM) with a formal means for efficient design exploration around a given Pareto design. It has two main components. The first one is an Interactive MultiObjective Optimization Procedure (IMOOP) which provides designers with a Pareto surface approximation at a given Pareto design. This Pareto surface approximation is used for efficient design exploration. A formal decision making strategy called the Interactive Decision Making Strategy (IDMS) is also developed which includes class functions for capturing the DM's local preferences. In order to address multidisciplinary systems design, the iMOODS is modified to provide for Multidisciplinary Design Optimization (MDO). A response surface based Concurrent SubSpace Optimization (CSSO) algorithm is implemented within iMOODS to address the computational requirements, increased complexity and organizational challenges encountered in multidisciplinary systems design and optimization. Each of the strategies has been tested successfully using a number of multiobjective design test problems. The results indicate that the DM's preferences can easily be captured and the Pareto points that reflect these preferences can be generated efficiently using the Pareto surface approximation. This is important for the confidence of the DM when using an interactive framework such as iMOODS for obtaining a satisfactory final Pareto design in a minimal number of iterations. The results also indicate that the iMOODS with MDO capabilities is suitable for multidisciplinary systems design.