TY - GEN
T1 - A simultaneous computing framework for metamodel-based design optimization
AU - Long, Teng
AU - Wang, Lv
AU - Wu, Di
AU - Guo, Xiaosong
AU - Liu, Li
N1 - Publisher Copyright:
Copyright © 2014 by ASME.
PY - 2014
Y1 - 2014
N2 - At the aim of reducing the computational time of engineering design optimization problems using metamodeling technologies, we developed a flexible distributed framework independent of any third-part parallel computing software to implement simultaneous sampling during metamodel-based design optimization procedures. In this paper, the idea and implementation of hardware configuration, software structure, the main functional modular and interfaces of this framework are represented in detail. The proposed framework is capable of integrating black-box functions and legacy software for analyzing and common MBDO methods for space exploring. In addition, a message-based communication infrastructure based on TCP/IP protocol is developed for distributed data exchange. The Client/Server architecture and computing budget allocation algorithm considering software dependency enable samples to be effectively allocated to the distributed computing nodes for simultaneous execution, which gives rise to decreasing the elapsed time and improving MBDO's efficiency. Through testing on several numerical benchmark problems, the favorable results demonstrate that the proposal framework can evidently save the computational time, and is practical for engineering MBDO problems.1
AB - At the aim of reducing the computational time of engineering design optimization problems using metamodeling technologies, we developed a flexible distributed framework independent of any third-part parallel computing software to implement simultaneous sampling during metamodel-based design optimization procedures. In this paper, the idea and implementation of hardware configuration, software structure, the main functional modular and interfaces of this framework are represented in detail. The proposed framework is capable of integrating black-box functions and legacy software for analyzing and common MBDO methods for space exploring. In addition, a message-based communication infrastructure based on TCP/IP protocol is developed for distributed data exchange. The Client/Server architecture and computing budget allocation algorithm considering software dependency enable samples to be effectively allocated to the distributed computing nodes for simultaneous execution, which gives rise to decreasing the elapsed time and improving MBDO's efficiency. Through testing on several numerical benchmark problems, the favorable results demonstrate that the proposal framework can evidently save the computational time, and is practical for engineering MBDO problems.1
UR - http://www.scopus.com/inward/record.url?scp=84961361893&partnerID=8YFLogxK
U2 - 10.1115/DETC201434594
DO - 10.1115/DETC201434594
M3 - Conference contribution
AN - SCOPUS:84961361893
T3 - Proceedings of the ASME Design Engineering Technical Conference
BT - 40th Design Automation Conference
PB - American Society of Mechanical Engineers (ASME)
T2 - ASME 2014 International Design Engineering Technical Conferences and Computers and Information in Engineering Conference, IDETC/CIE 2014
Y2 - 17 August 2014 through 20 August 2014
ER -