Case Study 4: Ink Delamination Case Study Solutions
Define
GHF specializes in the printed graphic business. One of GHF’s largest customers has reported that the ink on the grid of one of their printed products was delaminating, causing the client’s production line to stop. A Six Sigma team was formed in order to determine the root cause of the problem. The purpose of this report is to improve the delamination rating to customer approved levels.
Measure
The Key Process Output variable is the number of delamination defects. The current state for delamination errors is 200,000 DPM (25% defective). The team decided that two studies would be made to narrow the search for root causes of the problem:
- Study 1. Three inspectors rated the delamination of the same thirty parts. This test was conducted to test the ability of the inspectors to accurately measure the delamination as compared to a master part and to each other.
- Study 2. A DOE experiment was performed (26-2) using potential key input variables that had been recognized by the team in a brainstorming session. These were: belt speed, jet oven temperature, pot life, oven temperature, oven time and mesh.
Analyze
Study 1
A two-way ANOVA was performed on the delamination rating data by inspector and by part. The results were as follows:
Two-way ANOVA: Delamination versus Parts, Appraiser
Analysis of Variance for Delamina
Source DF SS MS F P
Parts 29 1002.367 34.564 98.16 0.000
Appraisers 3 9.367 3.122 8.87 0.000
Error 87 30.633 0.352
Total 119 1042.367
Individual 95% CI
Appraiser Mean --+------+------+------+------
Inspector1 5.67 (------*------)
Inspector2 4.97 (------*------)
Inspector3 5.00 (------*------)
Master 5.23 (------*------)
--+------+------+------+------
4.80 5.10 5.40 5.70
As can be seen from the Minitab output, Inspector 1 is giving significantly different ratings than the Master and the other operators. If the difference between operators is a widespread phenomenon, this can be a factor in GHF shipping out defective parts that have obtained a better than expected rating. If we re-fit our ratings without inspector 1, we may notice that there is no statistically significant difference in ratings between the two remaining operators and the master part.
Two-way ANOVA: score versus part and appraiser
Analysis of Variance for score
Source DF SS MS F P
part 29 759.600 26.193 103.11 0.000
Appraisers 2 1.267 0.633 2.49 0.091
Error 58 14.733 0.254
Total 89 775.600
Individual 95% CI
Appraisers Mean --+------+------+------+------
Inspector2 4.967 (------*------)
Inspector3 5.000 (------*------)
Master 5.233 (------*------)
--+------+------+------+------
4.800 4.960 5.120 5.280
Although we recognize some operator training is needed, we believe that our rating system has the potential to produce repeatable results.
Study 2
A DOE experiment was carried out by the team to study 6 factors. We used a 26-2 experiment which contains 16 unique run combinations. For each of these run combinations, we collected 3 part samples (or replications). Note: for this experiment, the cost of setup for a particular run combination is much greater than generating multiple samples once the process configuration is set.
The results of the data analysis were as follows:
Fractional Factorial Fit: Delamination versus Belt Speed, Jet Oven Tem, ...
Estimated Effects and Coefficients for Delamina (coded units)
Term Effect Coef SE Coef T P
Constant 6.500 0.2946 22.06 0.000
Belt Spe -0.833 -0.417 0.2946 -1.41 0.167
Jet Oven -0.000 -0.000 0.2946 -0.00 1.000
Pot Life -1.417 -0.708 0.2946 -2.40 0.022
Oven Tem -2.750 -1.375 0.2946 -4.67 0.000
Oven Tim -1.583 -0.792 0.2946 -2.69 0.011
Mesh 0.000 0.000 0.2946 0.00 1.000
Belt Spe*Jet Oven -0.167 -0.083 0.2946 -0.28 0.779
Belt Spe*Pot Life 0.250 0.125 0.2946 0.42 0.674
Belt Spe*Oven Tem -0.083 -0.042 0.2946 -0.14 0.888
Belt Spe*Oven Tim -0.250 -0.125 0.2946 -0.42 0.674
Belt Spe*Mesh -0.500 -0.250 0.2946 -0.85 0.402
Jet Oven*Oven Tem 1.250 0.625 0.2946 2.12 0.042
Jet Oven*Mesh -1.167 -0.583 0.2946 -1.98 0.056
Belt Spe*Jet Oven*Oven Tem -0.083 -0.042 0.2946 -0.14 0.888
Belt Spe*Jet Oven*Mesh 0.333 0.167 0.2946 0.57 0.576
Analysis of Variance for Delamina (coded units)
Source DF Seq SS Adj SS Adj MS F P
Main Effects 6 153.250 153.250 25.5417 6.13 0.000
2-Way Interactions 7 40.000 40.000 5.7143 1.37 0.251
3-Way Interactions 2 1.417 1.417 0.7083 0.17 0.844
Residual Error 32 133.333 133.333 4.1667
Pure Error 32 133.333 133.333 4.1667
Total 47 328.000
As can be seen from the first DOE experiment, the significant main effects for an alpha error level of 0.05 are Pot life, Oven Temperature and Oven Time. The only significant two way interaction is Jet Oven*Oven Temperature.
After removing the insignificant terms, we further analyzed the same data set considering only significant (either as a main effect or included in an interaction). The results are shown below:
Fractional Factorial Fit: Delamination versus Jet Oven Tem, Pot Life, ...
Estimated Effects and Coefficients for Delamina (coded units)
Term Effect Coef SE Coef T P
Constant 6.500 0.2855 22.77 0.000
Jet Oven -0.000 -0.000 0.2855 -0.00 1.000
Pot Life -1.417 -0.708 0.2855 -2.48 0.017
Oven Tem -2.750 -1.375 0.2855 -4.82 0.000
Oven Tim -1.583 -0.792 0.2855 -2.77 0.008
Jet Oven*Oven Tem 1.250 0.625 0.2855 2.19 0.034
Analysis of Variance for Delamina (coded units)
Source DF Seq SS Adj SS Adj MS F P
Main Effects 4 144.92 144.92 36.229 9.26 0.000
2-Way Interactions 1 18.75 18.75 18.750 4.79 0.034
Residual Error 42 164.33 164.33 3.913
Lack of Fit 10 31.00 31.00 3.100 0.74 0.679
Pure Error 32 133.33 133.33 4.167
Total 47 328.00
As can be seen, all of the variables remain significant either as a main effect or as an interaction.
Using the response optimizer option in Minitab, the target level of 9 delamination ratingcould be achieved by using the following settings:
Jet Oven / MediumPot Life / 2.52 Hrs
Oven Time / 2 Hrs
Oven Temperature / 140 F
However, since GHF wants to minimize scrap costs, we desire a Pot Life to be as large as possible in order to minimize scrap costs.
After exploring the different configurations, the team chose the following settings, which also maintain a delamination rating of 9 with the highest possible Pot Life (hence reducing scrap costs):
Jet Oven / LowPot Life / 3.35 Hrs
Oven Time / 2 Hrs
Oven Temperature / 140 F
Improve
Based on the team’s findings, the following recommendations were made:
1)Increase operator training in order to minimize the discrepancies in delamination rating.
2)Change the process settings to:
Jet Oven / LowPot Life / 3.35 Hrs
Oven Time / 2 Hrs
Oven Temperature / 140 F
Since the belt speed and mesh variables are not significant to the delamination rating, they may be set at either the low or high setting. (Note: we may further explore setting the belt speed at the high setting if it can result in a lower cycle time. These recommendations are made under the assumption that there are no prohibitive costs, or adverse affects to other outputs by running at these settings.)
A simulation using the previous process settings was run to estimate the future state of the process. Based on 30 simulated runs, the capability of the process future state was calculated:
As it can be seen from the graph, the DPM is expected to reduce drastically.The expected DPM is only 20, with no current observed defects.
Control
It should be ensured that the process setting recommendations are incorporated into standardized setup procedure. With each new batch setup, employees are expected to complete a check list signifying that all key process input settings are appropriate. In addition, the manufacturer should monitor the process by sampling at the start and end of each batch to insure conformance to specification.