THE EFFECTS OF BUFFER PATTERNS ON THROUGHPUT IN CONWIP FLOW LINES: A SIMULATION STUDY
Sunitiya Thuannadee
Received: May 21, 2007; Revised: Sept 4, 2007; Accepted: Sept 10, 2007

Abstract
This research was a simulation experiment which aimed to investigate the effects of buffer patterns on the throughput of lines controlled by a CONWIP release mechanism. The hypothetical lines simulated were stochastic 11-station flow lines with the middle station as the bottleneck. ANOVA tests were used for the analysis of the simulation output, followed by Duncan’s Multiple Range tests. There were 6 buffer patterns which were distinguished by the amount of total buffer capacity and the buffer allocation approaches. The experiment was done at three levels of protective capacity (PC); 0%, 20%, and 40%, and two levels of coefficient of variation (CV) of processing times; 0.05 and 0.50. When CV = 0.05, ANOVA tests revealed a significant difference (p < 0.01) among buffer patterns only for the lines with PC of 0%, whereas when PC = 20% and 40%, there were insignificant differences among buffer patterns. When CV = 0.50, ANOVA tests indicated significant differences (p < 0.01) among buffer patterns at all PC levels. The important suggestion is that appropriate allocation of limited buffer capacity, an addition of protective capacity, and a reduction of system variation could increase the throughput rates to the desired level. Keywords: Release mechanism, CONWIP, buffers, protective capacity, bottleneck

Introduction
An important factor for achieving target throughput rate of production lines is the method of controlling the release of new units into the production system called a release mechanism. One recent release mechanism which has been of interest to researchers is CONWIP, first introduced by Spearman et al. (1989). Under the CONWIP release mechanism, as a production unit is