Abstract
Many statistical experimental designs are too costly or require too much raw material to be feasible for lab-scale fiber spinning experiments. In this study a four-factor response surface design is presented to study the fiber spinning process in detail at the lab scale. The time, cost, and amount of raw material required to execute the proposed design are compared to the typical completely randomized 24 factorial design used in fiber spinning experiments and also to a standard four-factor response surface design. Sample fiber data as well as analysis from a typical statistical software package is provided to further demonstrate the differences between each design. By designating some treatment factors in the design as hard-to-change, split-plotting is used to reduce the time, cost, and amount of raw material required to complete the experiment. The proposed split-plot design is faster and less expensive than a typical factorial design and has the advantage of fitting a more complex second-order model to the system. When compared to a standard response surface design, the proposed split-plot design provides the same second-order modeling capabilities but reduces the cost of the experiment by 53%, the total time by 36%, and the amount of polymer required by 24%. Thus, a split-plot response surface design based on hard-to-change factors is recommended in lab-scale spinning.