The correct level of model complexity in semiconductor fab simulation — Lessons learned from practice

Abstract

It is well-known that semiconductor manufacturing is one of the most complex types of production. Consequently, making the right decisions in order to get the optimum output of a semiconductor fab is quite challenging. For that reason, sophisticated discrete event simulation studies may be seen as a kind of magic tool for fab managers to face the ever-increasing complexity. But building complex dynamic simulation models in order to solve complex problems is not always the best solution. This article gives an overview of the discussion about the correct level of complexity of discrete event simulation models and lists the main reasons for the creation of overcomplex models. To provide guidance for practitioners, lessons learned are derived from four "real world" use cases in which the correct level of complexity was a major issue.

Model simplification in manufacturing simulation – Review and framework

“…How to trace model components that are eligible for simplification, i.e., are of no significant relevance in capturing the causal connection between model inputs and outputs (Benjamin et al, 1998;Law et al, 1993;Robinson, 2008b). Basically, eligible components qualify by allowing for their omission, aggregation or substitution (Pegden, Shannon, & Sadowski, 1990;Rank et al, 2016). Omission of model components builds on the assumption that system elements represented by them have no significant influence on system performance.…”

“…Substitution entails replacing complex model components with simpler ones that approximate behavior of the former. Note that aggregation and substitution are related, with the latter being the result of repeated aggregation (Rank et al, 2016).…”

“…Essentially, rules identify components by studying their behavior. Often mentioned examples are shown in Table 2 (Rank et al, 2016;Robinson, 1994Robinson, , 2014.…”

Panel: Education on Simulation Model Simplification – Beyond Rules of Thumb

“… Demands on simulation study set-up: The demands for shorter simulation study lead times increase relevance of model simplification (Chwif et al 2006).  Use of simulation resources: Progress made in simulation software and computer hardware is often erroneously used as an argument to build more detailed models, thereby neglecting possibilities for model simplification (Salt 1993;Chwif et al 2000;Rank et al 2016).  Limited research on the topic: So far, model simplification does not seem to be on the research agenda of the simulation and modelling community.…”

“…Apart from a number of authors targeting issues in isolation, structural support seems to be only available within semiconductor manufacturing and -to a lesser extent -within the community of researchers targeting simulation conceptual modelling. Starting from the high complexity of semiconductor manufacturing systems, a group of researchers works on a body of knowledge within their domain (Rose 2000;Rose 2007;Rank et al 2016). Exploration of methods of model simplification has been identified as one of the research themes within the field of conceptual modelling (Van der Zee et al 2011).…”