* To whom correspondence should be addressed: janet.allen@ou.edu; 405-550-3969 Copyright © 2011 by ASME 1 Proceedings of IDETC/CIE 2011 ASME 2011 International Design Engineering Technical Conferences & Computers and Information in Engineering Conference August 28-31, 2011, Washington, DC, USA DECT2011-47655 MANAGING UNCERTAINTY IN MULTISCALE SYSTEMS VIA SIMULATION MODEL REFINEMENT Ayan Sinha The G.W. Woodruff School of Mechanical Engineering Georgia Institute of Technology, Atlanta, GA, USA Jitesh H. Panchal School of Mechanical and Materials Engineering Washington State University, Pullman, WA, USA Janet K. Allen * The School Industrial Engineering The University of Oklahoma, Norman, OK, USA Farrokh Mistree The School Aerospace and Mechanical Engineering The University of Oklahoma, Norman, OK, USA ABSTRACT The motivating question for this article is: ‘How should a system level designer allocate resources for auxiliary simulation model refinement while satisfying system level design objectives and ensuring robust process requirements in multiscale systems? Our approach consists of integrating: (i) a robust design method for multiscale systems (ii) an information economics based approach for quantifying the cost-benefit trade-off for mitigating uncertainty in simulation models. Specifically, the focus is on allocating resources for reducing model parameter uncertainty arising due to insufficient data from simulation models. A comprehensive multiscale design problem, the concurrent design of material and product is used for validation. The multiscale system is simulated with models at multiple length and time scales. The accuracy of the simulated performance is determined by the trade-off between computational cost for model refinement and the benefits of mitigated uncertainty from the refined models. System level designers can efficiently allocate resources for sequential simulation model refinement in multiscale systems using this approach. Keywords: Simulation-based multiscale systems design; robust design; information economics; 1. FRAME OF REFERENCE - CONCURRENT PRODUCT AND MATERIAL DESIGN The motivating question for this article is: ‘How should a system level designer allocate resources for auxiliary simulation model refinement while satisfying system level design objectives and ensuring robust process requirements in multiscale systems? The paradigm of concurrent design of materials and product entails tailoring materials to meet specific performance requirements. Design of material refers to controlling the microstructure and design of product implies meeting the performance requirements. Hierarchy exists over multiple length and time scale in the process-structure, structure-property and property-performance relationships, Figure 1. Hence, concurrent material and product design can be viewed as a multiscale design problem. Simulation-based concurrent material and product design has a number of challenges 1 : a) the presence of both reducible and irreducible uncertainty in hierarchical design chains, b) the presence of uncertainties within individual simulation models, c) propagation of uncertainties in interconnected models through multiple scales, d) evolving simulation models, resulting in multiple models of different fidelities at different points in a design process, e) significant model development and execution costs, necessitating judicious use of computational resources, and f) efforts to reduce computational cost by model simplification and hence increasing uncertainty. Thus, concurrent design of material and product is a representative example for managing uncertainty through simulation model refinement. We consider the design of an Autonomous Underwater Vehicle made of metal matrix composites. The vehicle has the following multifunctional requirements: 1  The safe depth of operation of the submersible should exceed 5000 meters and greater depth is better.  The submersible must operate for at least 10 hours without re- surfacing or recharging. Greater duration is better.  Given weight of vessel to be 80 kilograms and allowing as large a payload as is feasible, a representative limit for the weight of the outer shell of the submersible may not exceed 20 kilograms, and a lighter shell is preferred. The operating temperature of the submersible may not exceed 20 degrees Celsius to ensure safe operation of it electronic equipment. Metal matrix composites are strong, stiff light weight metal- based composites reinforced by a metal, ceramic or an organic compound 3 . A new category of materials known as in-situ composites have been developed, wherein the reinforcements are generated in a