Chapter 8. Improving a Polymer Manufacturing Process

The British company MoldMat Ltd. manufactures granulated white plastic at a plant in Britain and supplies it to a molding plant in Italy, where it is made into white garden chairs and tables. However, the molding process goes through intermittent phases when its product quality drops, leading to yield losses at both the polymer and the molding plants. When a crisis occurs, teams are formed to tackle the problem, but the problem usually disappears for no apparent reason.

After yet another mysterious crisis occurs and resolves itself, Carl Linton, a young engineer with black belt training in Six Sigma, is tasked to solve the problem once and for all. Together with a small project team, Carl identifies two characteristics (Ys) that are of paramount importance relative to quality and yield: the polymer's melt flow index (MFI) and its color index (CI).

Carl and his team reanalyze data collected by the most recent crisis team. Because suspected relationships between the two responses and eight process factors fail to reveal themselves in this analysis, Carl suspects that measurement variation may be clouding results. Consequently, Carl and his team conduct Measurement System Analysis (MSA) studies on the measured Ys and Xs. The problematic variables turn out to be MFI (one of the two Ys) and filler concentration (one of the Xs).

Once the repeatability and reproducibility issues for these two variables are addressed, Carl and his team gather new data. They initiate their analysis by visualizing the data one variable at a time and two variables at a time. Then, they proceed to modeling relationships using the screening platform. They develop useful models for MFI and CI that include terms that might otherwise have been overlooked had Carl not used the screening platform.

The profiler is used to optimize MFI and CI simultaneously. Using sound estimates of the expected variation in the Hot Xs, Carl simulates the expected distributions for MFI and CI at the optimal settings. This confirms that the parts per million (PPM) rate should be greatly reduced. After running some successful confirmation trials, the changes are implemented.

One and a half years later, not a single batch of white polymer has been rejected by the molding plant. The savings from rejected batches alone amount to about £750,000 per annum. Additionally, because there are now no processing restrictions on the molding plant, savings of £2,100,000 per annum are being realized by MoldMat's big customer. This, in turn, leads to increased sales for MoldMat. These savings came at very little cost, as project-related expenditures were minimal.

Carl's odyssey takes him and his team through all of the steps of the Visual Six Sigma Data Analysis Process. In particular, he engages in some interesting work involving MSAs and modeling using the screening platform. A list of platforms and options used by Carl is given in Exhibit 8.1. The data sets he uses can be found at http://support.sas.com/visualsixsigma. Please join Carl and his team as they solve a very tough but typical manufacturing problem.

Figure 8.1. Platforms and Options Illustrated in This Case Study