Addressing Multiple Responses Using Sequential Kriging Optimization

• Activity Number: 533
• Type: Contributed
• Date/Time: Wednesday, August 7, 2013 : 10:30 AM to 12:20 AM
• Sponsor: Section on Physical and Engineering Sciences
• Author(s): Sayak Roychowdhury*+ and Dr. Theodore T. Allen
• Companies: Dept. of ISE, The Ohio State University and The Ohio State University
• Keywords: Kriging meta-modeling ; Gaussian stochastic process ; Desirability functions ; Pareto Surfaces

Abstract

Sequential Kriging Optimization method is a method to optimize stochastic black box problems based on a Gaussian stochastic process meta-model. This method provides a global prediction of the objective function and a measure of uncertainty at every point. Often, a system has multiple responses with different objectives (e.g. maximize, target or minimize) and varying degrees of importance. Modeling each of the responses separately, creating predictions for these responses, and then putting the predictions into desirability formulas to predict overall desirability is one relevant option. Separate response modeling also supports the enumeration of Pareto surfaces. Another option is to create composite desirability values for all experimental runs and to use these as responses for modeling with a single meta-model. These options are compared in relation to computation and complexity costs and the resulting estimated solution quality. Automotive examples relating to discrete event simulation are used to illustrate the methods and comparison issues.