White Papers
A Web-based Collaborative Environment for Bi-Level Integrated System Synthesis
Hongman Kim, Scott Ragon, James Mullins, and Jaroslaw Sobieszczanski-Sobieski
Abstract
Bi-Level Integrated System Synthesis (BLISS) is a systematic approach to optimize the design of complex systems. One key advantage of the BLISS algorithm is that disciplinary groups can optimize sub-systems autonomously while staying in control of their analysis tools and models. This paper discusses a Web-based implementation of BLISS that delivers these capabilities. The Web architecture provides tools to share analysis models and distribute the computational load of expensive simulations. Problem definition, execution, and progress monitoring are performed through Web interfaces that allow disciplinary groups from distant sites to work together in the design optimization process.
Introduction
Bi-Level Integrated System Synthesis (BLISS) is a method for the natural decomposition of a system design into subsystem optimizations and a system optimization without the loss of important couplings and interactions1. The BLISS algorithm overcomes challenges inherent in optimizing complex systems such as aerospace vehicles where multidisciplinary interactions are prominent. When applied to multidisciplinary design optimization (MDO) problems, BLISS has multiple advantages: First, a complex optimization problem can be divided into smaller problems that can be solved more quickly and easily. Second, BLISS allows sub-system optimizations to be performed by disciplinary experts autonomously. Third, approximate surrogate models are used to reduce the total computational cost, and the subsystem problems can be computed in parallel, thus fully utilizing available computing resources.
The BLISS algorithm has been successfully demonstrated for aerospace vehicle design problems1,2,3. In particular, the authors created a prototype BLISS implementation to verify the BLISS algorithm4. The exercise also demonstrated the capabilities of the ModelCenter5 framework to build sophisticated MDO algorithms. BLISS was executed in a serial manner for the prototype exercise, and one of the key observations was that the next generation BLISS environment must be capable of supporting simultaneous multi-user execution. The achievement of this goal required development of new capabilities beyond those of existing MDO tools. This work is part of a research project to develop a Web-centric architecture called CenterLink6. Its goal is three-fold: first, development of coarse-grained parallel computing capabilities for design trade off studies; second, development of collaboration capabilities by sharing design data; third, implementation of an advanced MDO algorithm that can be easily deployed to groups of people. This paper focuses on the third goal and describes how to implement the BLISS algorithm on CenterLink. The key is to provide user interfaces to guide users step by step through the required design optimization tasks. The BLISS user interfaces are presented and the advantages of the Web-based approach are discussed.
Download The Complete Paper
(A one-time registration will be requested)
