Pareto Optimization of Physical Systems Using Tuned Computer Simulators

• Activity Number: 501
• Type: Contributed
• Date/Time: Wednesday, August 12, 2015 : 8:30 AM to 10:20 AM
• Sponsor: Section on Physical and Engineering Sciences
• Author(s): P.H. Allen Chen* and Thomas Santner and Angela Dean
• Companies: The Ohio State University and The Ohio State University and The Ohio State University
• Keywords: Computer Experiments ; Pareto Front ; Pareto Set

Abstract

A sequential method, based on a tuned simulator, is proposed for determining the Pareto Front of a multiple-output physical system. The data are comprised of a fixed number of runs from the physical system and a variable number of runs from a computer simulator of the physical system, where additional runs can be made from the simulator. The simulator inputs are assumed to encompass "control" inputs whose levels can be changed in both the physical system and the simulator, as well as "tuning parameters" that control the numerical solution characteristics of the algorithm used by the simulator. Each stage of the proposed sequential procedure consists of pairs of steps. Based on the data sampled, the first step estimates the tuning parameters of the code to improve the simulator accuracy. Then, guided by an improvement function, the second step selects a new control input site at which the computer simulator is to be run. Using the Hypervolume Indicator function to assess Pareto Front accuracy, the sequential procedure is shown to perform well through examples from the multiobjective optimization literature.