Roll variability analysis for non-statisticians

Abstract

Roll variability can be characterized by several different methods. Off line cross-direction (CD) scans of thickness or coat weight give a snapshot of CD variation over a small machine-direction (MD) range. Roll hardness profiles have long been used to evaluate web uniformity in wound roll form - particularly for flagging potential web bagginess. Roll hardness profiles are easy to measure, but a strong correlation to roll quality depends on having a somewhat stable CD thickness or coat weight profile and relatively low MD or time variation.

Summary statistics from in-line scanning gauges (mean, range and standard deviation) sample the entire roll, but don't describe the distribution or location of the variation. 2D contour plots are helpful for visualizing roll variation, but alone aren't enough to translate that variation into a useful specification.

Since specification limits for good web processing are often much tighter than limits for good end use function, making the best use of scanning gauge data is helpful for efficiently focusing process improvement efforts. Analysis of Variance (ANOVA) is able to evaluate the uniformity of a web property (thickness, coat weight) from an inline scanning gauge and partition the roll variance in the MD and CD directions.

This paper shows how the combination of ANOVA, control charting, and contour plotting can be used to both quantify and visualize roll variability within and between rolls in order to drive web uniformity improvement.