Semiconductor manufacturing: decision analysis for fab site selection

Abstract

A decision analysis model that facilitates the assessment of a multi-criteria decision from among a variety of competing alternatives is presented in this paper. The model is tested for sensitivity and found to be robust when one criterion is changed while others are held constant. The decision itself is related to locating a semiconductor manufacturing plant in one of many equally attractive and equally problematic sites worldwide. It is essentially a problem under conditions of near-certainty with no EMV or payback considerations by construction. The decision analysis described herein is a data-driven methodology with a two-step process where an algorithm is used to narrow down the choice of locations from a starting count of eight to a more manageable group of five. In the second step of the process, a decision model is defined and implemented. Hierarchical decision modelling is used in this study to disaggregate the complex decision into its component parts, to quantify and normalise judgements and to assign weights to the decision alternatives to obtain the proper result.