An empirical investigation of the two-moment approximation for production lines

Abstract

Abstract Since Muth (1973) it has generally been accepted that the throughput of production lines with variable processing times depends primarily on the first two moments of the processing time distribution. However, several recent studies have suggested that differences in higher moments may lead to significant differences in throughput. In order to shed light on this question, we examine the applicability of the two-moment approximation over a variety of types of lines (both serial and assembly) and a variety of processing time distributions. We find significant differences in throughput (on the order of 5%) even among commonly used probability distributions, and large differences (as large as 28%) among highly skewed and kurtotic distributions. We also find that the differences among distributions generally increase with the number of stations in the line and with the coefficient of variation of processing times.