Taylor Corp. charged our Lean Six Sigma project team with finding a way to reduce setup time and costs on 33 Kluge presses in operation at one of its stationery units. Taylor is the parent company of over 100 printing operations (including Carlson Craft, Fine Impressions, Schmidt Printing) producing a broad array of products

Traditionally, operators set up a printing job by conducting a series of informal experiments on the press. The project team decided to explore a more scientific experimentation process using the design of experiments (DOE) method.

Results yielded by using DOE led to a 5% reduction in printing costs by moving to a less expensive foil for that job. More importantly, increased process knowledge gained from the method, combined with other improvements, led to a 40% reduction in setup time over the last two years.

DOE is a structured, organized method used to determine the relationship between the different factors affecting a process and its output. It involves designing a set of experiments where relevant process factors are varied systematically. Experiment results are analyzed to identify optimal conditions and the factors that most influence the results—and those that do not. The method also uncovers the existence of interactions and synergies between factors.

In the past, operators at the printing company would try various settings of pressure, temperature and print speed until they produced samples that were subjectively judged to be appropriate. Many combinations were often required to find one that worked. The average set-up time across all presses was 113 minutes per job in the year prior to the project. This added up to over 4,000 hours of labor per year, or the equivalent of two full-time positions.

The project team used DOE to understand process factors and to determine optimal targets for those factors, so that operators could complete setups more quickly. A print job that was of large financial significance to the company, and typical of its work, was selected for study. Insights gained from the experiment would benefit other jobs.

The traditional method used to configure machinery was a “good” versus “bad” evaluation based on subjective observation, not stated criteria. Conversely, in a designed experiment process, variables are identified and measured.

The project team proposed building a measurement system to evaluate printing quality. A manager of the printing company developed an objective measurement system that used a scale from 1 to 5 to rate print samples on four quality characteristics, each representing a different potential problem: bottoming out die, cutting through, spotting and foil marks.

Five product samples were created to demonstrate and rate each characteristic. Prints generated during the experiment were visually compared to assign a rating. A Gage Repeatability and Reproducibility (R&R) study was conducted to verify the functionality of the new measurement system. Gage R&R is a statistical tool used to measure the amount of variation in a measurement system arising from the measurement device and the people taking the measurement.

The samples were provided to various operators without any identification. The operators were asked to rate the samples. The study verified the functionality of the new measurement system.

A screening experiment using Design-Expert software from Stat-Ease, Inc. was designed with two objectives:

1) Determine which factors play a statistically significant role in setting up the print job to produce a defect-free sample. Significant factors would be studied in more detail in follow-up experiments.

2) Settle a long-term organizational dispute about the need to use a higher-quality and more expensive raw material.

The team chose Design-Expert based on its user-friendliness, strong analysis tools and product support.

The Six Sigma project team decided to a conduct an experiment, where four factors would be studied in 16 job runs.

a) Foil type (standard quality vs. a higher quality, more expensive type)

b) Temperature (low temperature 250°F vs. high temperature at 300°F)

c) Pressure (low pressure with platen distance from chase to counter die of 1/16´´ vs. high pressure using a die cut platen with a distance of 1/8´´)

d) Press speed (slow at 5 cranks from baseline vs. fast at 25 cranks from baseline)

The factors were measured against the four potential problems of bottoming out die, cutting through, spotting and foil marks.

All four of the response models were statistically significant with p-values well below the typical .05 threshold. A sample of the ANOVA (analysis of variance) analysis for one of the response variables is shown on p.21.

Models for two of the responses—bottoming out die and spotting—proved to have a very high statistical significance with R-Squared values above 90%. R-Squared, also called the coefficient of determination, is the proportion of variability in the data that is accounted for by the statistical model. The R-Squared value for cutting through was 39%, and for foil marks it was 26%. Pressure was the dominant factor in the models for three of the four responses: bottoming out die, cutting through and spotting. In the model for foil marks, pressure and press speed were both important.

The screening experiment answered the question about the need to use the expensive foil. Foil was statistically insignificant in all four models, meaning that the cheaper foil worked as well as the expensive one. This was exciting news as using the cheaper foil would be a source of significant savings.

A response surface study (CCD) was conducted to confirm results from the screening experiment and to check for non-linear effects. The study focused on the same factors used in the screening study minus foil type. Factor levels were adjusted slightly. Even though temperature had not proved statistically significant in the screening experiment, organizational wisdom suggested that temperature was still an important factor. It was therefore included again in the experiment. The experiment consisted of 20 runs including several replicates in order to more accurately assess experimental error. The foil marks response was also eliminated.

All three of the models proved statistically significant. The R-Squared was 86% for bottoming out die, 62% for cutting through and 83% for spotting. The models for bottoming out and spotting contained quadratic terms, while the cutting through model proved to be linear.

Lack of fit was insignificant in all three models. The response surface analysis showed once again that pressure was by far the most important factor studied with optimal results achieved at fairly high pressure. Temperature and speed had less significance. These were desirable results for a number of reasons. Running at higher speeds increases productivity. Flexibility in temperature means that less time needs to be spent waiting for the press to reach very specific temperatures.

The perturbation plot (p.22) conveys information about factor control in press setup in graphical form by showing how the response changes for each factor—pressure, temperature and press speed—alone, while holding all others constant. It shows that pressure, Factor A, is the variable that must be closely controlled. The steep slope for pressure suggests that small changes in the pressure level have dramatic impact on overall desirability. The slopes of the plots for temperature and press speed (factors B and C, respectively) are more horizontal, suggesting that changes in level have less significant impact on overall desirability.

The strong statistical significance of pressure shown throughout all of the experimentation supported a significant investment decision that was being evaluated by the organization. The company was considering the purchase of an adjustable impression device that changes pressure by shortening or lengthening the side arms. The importance of pressure adjustment revealed by the experiments helped tip the scales in the decision-making process. The understanding of the sensitivities provided by DOE and the investment in this adjustable impression device were key factors in reducing average setup times by 40%, to about 70 minutes per job.

The project team felt the DOE studies resulted in tangible benefits. The experiments were easy to conduct and provided useful insights into how to improve the setup process. DOE can be a useful tool for process improvement across the printing industry.

ONLINE: graphicartsonline.com/management