Scheller input the factors into the DOE software and generated a 16-run experiment in which each of the eight design points were replicated as shown in Figure 1. The order is randomized to reduce bias from lurking time-related variables such as change in ambient temperature, or humidity and the like. The results were then entered into the software and analyzed.
Figure 2: Meltability model graph
First, the meltability response was considered. A meltability score within the range of 3 to 6 was desired for this trial product. Figure 2 shows a graph of the model for meltability. The model has an F-value of 14.99, which indicates there is only a 0.05% chance (p = 0.0005) these results could be caused by noise. “The graph shows that, for this experiment, when the Emulsifier Type 1 is used, meltability is consistently within specification,” Scheller says. “On the other hand, when Emulsifier Type 2 is used, meltability is all over the map.” The bars indicate the least significant difference (LSD) for 95% confidence. These show a clear separation in meltability by Type, see emulsifier ratio D, for example.
Figure 3: pH model graph for Type 1 emulsifier