Some alternative methods for monitoring multiple-stream processes

Abstract

An example of a "multiple-stream process" is a multiple-head machine which fills bottles to a common level. Instead of setting up a control chart for each head, it is convenient to monitor all of the heads simultaneously on a single control chart. Various methods mentioned in the literature about this topic are discussed. The problem with the conventional method is discussed and two new methods which can be used to monitor multiple-stream processes are proposed. Four methods are discussed in detail. The first method is based on the distribution of the maximum and minimum values in a sample selected from the process. The second method uses correction factors to widen the control limits. The third method is the conventional technique used to monitor these types of processes. The fourth method is an empirical method based upon the results of a Monte Carlo computer simulation. The average run lengths of the four methods are compared and the methods are applied to real-life data sets to see how they perform in practice.