1. Problem Statement:
Color Concentrate usage is above the
standard on 500 - 700 ton injection molded
parts, resulting in a raw material variance.
Project Definition
CTC:
Accurate use of concentrate.
CTQ:
Maintain accurate color match and plastic
stability.
Defect Definition:
Concentrate usage that is not within the
standard of 1.85%-2.15%.
Project Objective:
Develop a process to accurately measure 2%
concentrate into the molding process, by
May 1, 2001.
Benefits:
•Annual cost save of $292,184.
•Consistent color match.
•Part Stability.
Define
DPMO, Short Term Sigma and RTY:
Baseline 1,000,000 -3.00 0.0%
Goal 200,000 2.34 80.0%
Current 43,875 4.17 90.0%

2. Baseline Performance Measure
Comments:
•30 Samples Collected By
S.VanMetre During 1 Week
Over 3 Shifts
•Data is long term.
Concentrate Usage Percentage
0.00%
1.00%
2.00%
3.00%
4.00%
5.00%
6.00%
7.00%
8.00%
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29
# of Samples
Percentage
Actual Usage Machine 67
Average Usage #67
Standard Usage
Average of
4% Over
Standard
Measure

3. 2% StandardBaseline
7
6
5
4
3
2
1
0
Boxplots of Old and Standard
Ho:Baseline = 2% Standard
Ha:Baseline ≠ 2% Standard
N Mean StDev SE Mean
Baseline 50 5.710 0.605 0.086
2% Stnd 50 2.12 1.34 0.19
Difference = mu Mach 67 - mu Mach 65
Estimate for difference: 3.593
95% CI for difference: (3.178, 4.007)
T-Test of difference = 0 (vs not =): T-Value = 17.30 P-Value = 0.000 DF = 68
Old Equip. at Standard
is not better due to
standard deviation.
The deviation could
allow the percentage
to be too low.
T - Test #1
Improve
Test for Equal Variance
Test Statistic: 17.599
P-Value : 0.000
Baseline
vs
2% Standard
2 Sample T Test

4. T - Test #2
Improve
Ho:Baseline = New Equip at 2%
Ha:Baseline ≠ New Equip at 2%
N Mean StDev SE Mean
Old System 50 5.710 0.605 0.086
New System 50 1.8833 0.0855 0.012
Difference = mu Old - mu New
Estimate for difference: 3.8270
95% CI for difference: (3.6535, 4.0005)
T-Test of difference = 0 (vs not =): T-Value = 44.30 P-Value = 0.000 DF = 50
The new blender, dispenses
the concentrate more accurately
than the old system.
The Average amount of
concentrate used per mix, is
improve by nearly 4%.
2 Sample T Test
New at 2%Baseline
7
6
5
4
3
2
1
Boxplots of Old and New
Test for Equal Variance
Test Statistic: 36.364
P-Value : 0.000
Baseline Equip.
vs
New Blender

5. T - Test #3
2 Sample T Test New vs Mat’l Order Improve
Ho:Conc 2nd = Conc. 1st
Ha:Conc 2nd ≠ Conc. 1st
N Mean StDev SE Mean
Conc 2nd 50 1.8833 0.0855 0.012
Conc 1st 50 1.9377 0.0513 0.0073
Difference = mu Conc 2nd - mu Conc 1st
Estimate for difference: -0.0543
95% CI for difference: (-0.0824, -0.0263)
T-Test of difference = 0 (vs not =): T-Value = -3.85 P-Value = 0.000 DF =
80
Conc 2nd Conc 1st
1.5
1.6
1.7
1.8
1.9
2.0
2.1
Boxplots of Conc 2nd and Conc 1st
2 Sample T Test
Dispensing Concentrate 1st,
has a smaller standard deviation.
The blender will dispense
a more accurate percentage of
concentrate.
Test for Equal Variance
Test Statistic: 5.700
P-Value : 0.019
New Equipment Dispensing
Concentrate 2nd
vs
Concentrate 1st

6. Improve
Concentrate Usage by Blender
0.00%
1.00%
2.00%
3.00%
4.00%
5.00%
6.00%
7.00%
8.00%
1
3
5
7
9
11
13
15
17
19
21
23
25
27
29
31
33
35
37
39
41
43
45
47
49
Samples
Percentage
Goal
Old System
Concentrate 1st
Current Performance Measure
Old System
Average of
4% Over
Standard
Concentrate
Dispensed
1st.

7. Improve
Concentrate Usage by Blender
0.00%
1.00%
2.00%
3.00%
4.00%
5.00%
6.00%
7.00%
8.00%
1
3
5
7
9
11
13
15
17
19
21
23
25
27
29
31
33
35
37
39
41
43
45
47
49
Samples
Percentage
Goal
Old System
Concentrate 1st
Current Performance Measure
Old System
Average of
4% Over
Standard
Concentrate
Dispensed
1st.