다분야 통합 최적설계를 위한 분산컴퓨팅 시스템 개발 = Development of MDO-Enabling Distributed Computing System

국문 초록 (Abstract)

영문초록:MDO is an engineering design methodology for making a globally optimal decision among a number of different disciplines (for example, structure, aerodynamics, thermodynamics, and reliability) whose local optimal decisions are often mutually conflicting. The MDO framework is a software environment which allows the user to integrate analysis programs, optimization programs, CAD systems, database systems, and GUIs into a distributed software system for engineering design applications on heterogeneous platforms. Major challenges for the development of a general-purpose MDO framework are: (1) support for legacy software and data on heterogeneous platforms, (2) integration of various engineering design activities including both computational tasks and human interactions, (3) and support for domain-specific design functionality. Until now, the IT research community has so far neglected these issues and there are no computing framework to address them effectively. In this thesis, we present a distributed computing framework intended to facilitate the efficient implementation of an MDO-based integrated design system for an engineering application. We call this framework MEDIC (MDO-Enabling DIstributed Computing framework). Main features of MEDIC are integration of legacy and commercial programs on heterogeneous platforms; convenient adding or deleting analysis and optimize resources; automated execution of problems and movement of data; and visualizing final optimization and analysis results. Our design approach to MEDIC is to propose a layered architecture which consists of distributed computing technology, distributed data management technology, system integration technology and program-specific technology. Each layer is designed as independently as possible so that advanced general purpose distributed middleware and domain-specific legacy software/data can be separately and flexibly managed and easily replaced. In the current implementation of MEDIC, we use a fault-tolerant Linda variant called Persistent Linda for the distributed computing technology layer (distributed middleware layer). We have developed an agent systems to wrap legacy software and data and to provide standard interfaces to distributed middleware layer. The scheduling and execution of MDO-based design activities are implemented based on the Linda Tuple Space model. For this research, we have developed a prototype system of MEDIC and applied it for building an integrated design environment for super-high temperature vacuum furnaces