1 / 3

Design of Experiments and Variable Screening in Large-Scale Models

Design of Experiments and Variable Screening in Large-Scale Models. Jorge L. Romeu (1) and John J. Salerno (2) First International Workshop on Social Computing, Behavioral Modeling, and Prediction Phoenix, Arizona; April 1-2, 2007.

Ava
Télécharger la présentation

Design of Experiments and Variable Screening in Large-Scale Models

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Design of Experiments and Variable Screening in Large-Scale Models Jorge L. Romeu (1) and John J. Salerno (2) First International Workshop on Social Computing, Behavioral Modeling, and Prediction Phoenix, Arizona; April 1-2, 2007 (1) Department of Mechanical and Aerospace Engineering, Syracuse University. (2) Air Force Research Laboratory (AFOSR/RIEA), Rome Research Site, Rome NY

  2. Problem Statement Proposed Solution Current Situation: Given NOEM simulation model of a nation-state and military situation. Given some Dependent variables of interest known as Responses. Given a Large set of Independent variables known as Main Factors. Find the relationship between them and build a reduced Meta Model. Screen Meta Models and identify the Best Main Factors and Interactions. Select those Meta Models whose Key Factors are within the user “domain of action”. Use such Meta Models for analysisofoperations, and optimization. Need for Screening NOEM Simulation Model currently has far too many Factors or Variables. Which problem Factors have the most impact on the Response of interest? Which of the factor Interactions, also impact the Response of interest? What is the Sign and Magnitude of each of these factors and interactions? What Functional Form (linear, quadratic) does the Response function have? How much Variation do they Capture? What is the percentage Explanation?

  3. Current State of the Research DOE Analysis Methods Attempted • Full and Fractional Factorial Designs • Plackett-Burnam Experimental Designs • Latin Hypercube Sampling Approach • Response Surface Methodology • Multiple Regression Approach • Controlled Bifurcation Method • Hierarchical and Bayesian Approaches And we are still searching and assessing other DOE methods Main Problems Encountered • Large Interaction between Main Factors and Interactions • This creates Factor Signs and Significance Instability • Masking the Screening of Key Problem Factors

More Related