Reducing Label Nonconformances by DOE and Logistic Regression


Artiles-Leon, Noel   (1993, ASQC)   University of Puerto Rico; Mayaguez, PR

Annual Quality Congress, Boston MA    Vol. 47    No. 0
QICID: 9954    May 1993    pp. 174-180
List $10.00
Member $5.00

This article is not available online. Contact us to receive a scan of the archive, in PDF format.

Article Abstract

What affects the probability that a container label will be attached improperly? A logistic regression analysis examined the following factors: speed of the conveyor carrying the containers to the labeling station; the type of liner (two were available) for the labels; the length of the protruding edge of the labels; the horizontal angle of the sensor that detected an arriving container; and the width of the track that carried the containers. An experimental design reduced the values of these factors into 16 experimental states. For each state the experiment examined 100 labeling events and recorded the percent of defects. To predict and minimize the proportion of defects, the results suggested four models: weighted least squares; weighted least squares without the one state in which 99% of the labels were defective; maximum likelihood; and arcsine. All four models produced similar predictions and suggested that the most important factors to control were the sensor angle and track width. After making these adjustments, the defect rate decreased from 5%-10% to 0.96%.


Defects,Design of experiments (DOE),Process improvement,Regression analysis

Browse QIC Articles Chronologically:     Previous Article     Next Article

New Search

Featured advertisers

ASQ is a global community of people passionate about quality, who use the tools, their ideas and expertise to make our world work better. ASQ: The Global Voice of Quality.