Multiple response optimization in a fully automated FAB: an integrated tool and vehicle dispatching strategy

  • Authors:

  • Jonah C. Tyan;Timon C. Du;James C. Chen;Ir.-Hui Chang

  • Affiliations:

  • Taiwan Semiconductor Manufacturing Company, 6 Creation Rd., Sec. 2, Hsinchu 300, Taiwan, ROC;Department of Decision Sciences and Managerial Economics, The Chinese University of Hong Kong, Shatin, New Territory, Hong Kong, China;Department of Industrial Engineering, Chung Yuan Christian University, 22 Pu-Jen, Pu-Chung Li Chung-Li 320, Taiwan, ROC;Department of Manufacturing Engineering, National Cheng-Kung University, 1 University Road, Tainan 600, Taiwan, ROC

  • Venue:
  • Computers and Industrial Engineering
  • Year:
  • 2004

Quantified Score

Hi-index 0.00

Visualization

Abstract

In today's tight, competitive, and volatile market in semiconductor manufacturing, a successful semiconductor manufacturer has to look into multiple performance indices such as cycle time and on-time delivery. Effective dispatching mechanism is a widely used strategy to achieve these performance indices simultaneously. In the past, dispatching studies in semiconductor manufacturing have primarily focused on the level of tool dispatching. This paper presents an integrated tool and vehicle (ITV) dispatching strategy to consider multiple performance measures in a fully automated fab environment. The ITV dispatching strategy was developed using a state-dependent methodology and multiple response optimization. In order to build a simulation-based automated fab, an integrated modeling approach was proposed to automate both the manufacturing process and the automated material handling system. A case study based on a local fab is described to examine the performance impact of the ITV dispatching rule measured by cycle time, work-in-process, on-time delivery, and lot delivery time. The results of the simulation experiments and analysis show that the ITV dispatching rule is superior to the use of a static dispatching rule, consisting of an average of 15% improvement for on-time delivery and 5% for other performance measures. Furthermore, the proposed modeling framework features high-fidelity, real-time operations and re-configuration; and it can be easily used in other applications.