You are on page 1of 9

Lean Six Sigma Improve Phase

At the outset of this project we looked at the high level process metric of raw material yield as this was the area of concern for the organisation. We established that Line 2 had a raw material yield score of 99.1% set against a target of 99.8%. From a review of the financial performance metrics for the full year 2012 we determined that each .1% represented approximately 15,000. If we could improve Line 2 performance to the target value this represented a potential gain / saving in the order of 105,000 and we had set ourselves a target of achieving 50% of this figures. In the course of the Define, Measure and Analyse phases we have come to better understand our process, we have identified what we believe to be the main drivers of process variation and improving these would be the key to delivering a positive process shift. As we observed during the Measure phase our process was neither in statistical control or capable of producing output consistently within specification. Indeed our process performance exercise indicated that two thirds of the output would be outside the upper specification limit. While the team felt that they had identified the main sources of process variation it was by no means certain that the course of action we had proposed to the project sponsor would deliver the desired outcome. Our strategy was simple improve the machine, method and man. It was agreed that Standard Operating Procedures would be updated. As part of this exercise the senior filler operator and shift supervisor were tasked with developing best practice in a number of operational areas including start-up, close down and monitoring. These new / revised practices would form the basis of the training documentation. In addition to the annual line overhaul a new set of fill tubes were purchased and installed as it was noted that the fill volume range was too dispersed and it was critical that this was addressed if the operator was going to be in a position to exercise any control over the process output. I have included below the original X-bar & R control chart that was developed during the measure stage just as a visual reminder as to how unstable the process was at that point.

Line 2 Fill Volume Control Chart


2020
1 1 1 1 1 1 1

UCL=2016.78 _ _ X=2006.10

Sample Mean

2010

2000
1 1

LCL=1995.43
1

1990 1 17 33

1 1

1 1 1

49

65

81 Sample

97

113

129

145

60
1 1

1 1

Sample Range

40

UCL=39.14 _ R=18.51

20

0 1 17 33 49 65 81 Sample 97 113 129 145

LCL=0

5 samples taken at 2 hour interval 2 shift cycle for 4 weeks Reference period w/b 14/01/13 - w/e 08/02/13

The original X-bar & R control chart developed in the Measure phase The moment of truth!

Line 2 Fill Volumes Post Improvement


2008 UCL=2007.21

Sample Mean

2004

_ _ X=2002.53

2000 LCL=1997.84 1 9 17 25 33 41 Sample 49 57 65 73

UCL=17.19 15

Sample Range

10

_ R=8.13

5 0 1 9 17 25 33 41 Sample 49 57 65 73

LCL=0

5 samples taken at 1 hour intervals Reference period 19/03/2013 - 25/03/2013 inclusive

X-bar & R Control chart Post Improvement

Having completed the overhaul we needed to measure the process, please see above the newly developed X-bar & R control chart for line 2. While it is only one element of the improvement initiative it is extremely important and I am delighted to report that the process is approaching statistical control with no data points being observed outside of the process determined control limits. There are a number of patterns that warrant investigation, we can observe a run of eight consecutive points above the mean. I suspect this can be explained by the operator trying to influence the process and will be tracked going forward.

Before/After Capability Comparison for C7, ... vs C1, ... Process Performance Report
Capability Histogram Are the data inside the limits and close to the target?
LSL TargetUSL

Process Characterization Before Total N Subgroup size Mean StDev (overall) StDev (within) 800 5 2006.1 9.5329 7.9579 After 400 5 2002.5 3.7130 3.4942 Change

Before

-3.5760 -5.8199 -4.4638

Capability Statistics Before After 1.57 0.22 0.67 26.50 25.26 265000 252650 1.67 0.24 0.71 23.95 239502 Change 0.96 0.26 0.78 -29.88 -29.35 -298750 -293460 0.94 0.28 0.85 -31.56 -315591 Actual (overall) Pp Ppk Z.Bench % Out of spec (obs) % Out of spec (exp) PPM (DPMO) (obs) PPM (DPMO) (exp) Potential (within) Cp Cpk Z.Bench % Out of spec (exp) PPM (DPMO) (exp)

After

0.61 -0.04 -0.12 56.38 54.61 563750 546109 0.73 -0.05 -0.14 55.51 555094

1970

1980

1990

2000

2010

2020

2030

2040

Actual (overall) capability is what the customer experiences. Potential (within) capability is what could be achieved if process shifts and drifts were eliminated.

Before & After Capability Comparison Report As this report suggests it looks at the process output data before and after the improvement initiative in both numerical and graphical terms. The process characteristics have improved substantially with the mean being reduced by 3.58mls toward the target value. The process standard deviation has also been reduced which indicates improved stability and consistency in the process. The capability statistics really tell the story with an impressive improvement in the Pp, process performance index (process spread) now being reported at 1.57. However the Ppk index (process centre) of 0.26 is still an issue, Ppk measures the process centre in relation to the upper and lower specification limits. The PPM (parts per million) suggests that for every 1,000,000 units produced 239,502 will be filled beyond the 2005 upper specification limit.

We can see this clearly from the graphic that we still have a substantial quantity of output outside the upper specification limit. While this is not an issue for the end user it does represent a considerable giveaway. In an ideal world we would like our process mean to equal or nominal or target value. The elephant in the room While the improvement in process spread was substantial (Pp 0.61 1.57) it still presented a problem. If our target value is set at 2000mls and the process Standard Deviation is 3.713 how can we hope to achieve a situation where we produce all output within the 2005 upper specification limit. The answer to this question was simple under the existing conditions we couldnt as the natural variation of the process exceeded this tolerance. However this would not prevent us from making additional progress by optimising the operating conditions of our process. A theoretical exercise was conducted on the post improvement process data to see what effect a reduction in bowl pressure might have re. fill heights. If we reduced the bowl pressure thereby reducing the fill heights in 2mls increments what could we achieve? Simulations
Before/After Capability Comparison for C1, ... vs C13, ... Summary Report
Reduction in % Out of Spec 55% % Out of spec was reduced by 55% from 25.26% to 11.41%. Lower Spec 1970 Customer Requirements Target 2000 Upper Spec 2005

Was the process standard deviation reduced? 0 0.05 0.1 > 0.5

Statistics

Before 2002.5 3.7130 1.57 0.22 0.67 25.26 252650

After 2000.5 3.7130 1.57 0.40 1.20 11.41 114137

Change -2 0 0.00 0.18 0.54 -13.85 -138513

Yes

No
P = 0.500 Did the process mean change?

0.05

0.1

> 0.5

Yes
P = 0.000 Actual (overall) Capability Are the data inside the limits and close to the target?
LSL Target USL

No

Mean Standard deviation Capability Pp Ppk Z.Bench % Out of spec PPM (DPMO)

Comments Before: C1, ... After: C13, ...

Before

Conclusions -- The process standard deviation was not reduced significantly (p > 0.05). -- The process mean changed significantly. It is now closer to the target (p < 0.05). Actual (overall) capability is what the customer experiences.

After

1st iteration minus 2mls

This exercise demonstrated that by reducing the process uniformly by 2mls we could potentially reduce the out of specification output by 55%. It also had the added benefit of centralising the process.
Before/After Capability Comparison for C1, ... vs C19, ... Summary Report
Reduction in % Out of Spec 84% % Out of spec was reduced by 84% from 25.26% to 4.06%. Lower Spec 1970 Customer Requirements Target 2000 Upper Spec 2005

Was the process standard deviation reduced? 0 0.05 0.1 > 0.5

Statistics

Before 2002.5 3.7130 1.57 0.22 0.67 25.26 252650

After 1998.5 3.7130 1.57 0.58 1.74 4.06 40627

Change -4 0 0.00 0.36 1.08 -21.20 -212023

Yes

No
P = 0.500 Did the process mean change?

0.05

0.1

> 0.5

Yes
P = 0.000 Actual (overall) Capability Are the data inside the limits and close to the target?
LSL Target USL

No

Mean Standard deviation Capability Pp Ppk Z.Bench % Out of spec PPM (DPMO)

Comments Before: C1, ... After: C19, ...

Before

Conclusions -- The process standard deviation was not reduced significantly (p > 0.05). -- The process mean changed significantly. It is now closer to the target (p < 0.05). Actual (overall) capability is what the customer experiences.

After

2nd iteration minus 4mls Again this reduction had the potential to reduce the out of specification output by 84%. It also created a debate, we had a potential situation whereby the process mean was less that the target value but well within the packers regulations.

Before/After Capability Comparison for C1, ... vs C25, ... Summary Report
Reduction in % Out of Spec 96% % Out of spec was reduced by 96% from 25.26% to 1.12%. Lower Spec 1970 Customer Requirements Target 2000 Upper Spec 2005

Was the process standard deviation reduced? 0 0.05 0.1 > 0.5

Statistics

Before 2002.5 3.7130 1.57 0.22 0.67 25.26 252650

After 1996.5 3.7130 1.57 0.76 2.28 1.12 11242

Change -6 0 0.00 0.54 1.62 -24.14 -241408

Yes

No
P = 0.500 Did the process mean change?

0.05

0.1

> 0.5

Yes
P = 0.000 Actual (overall) Capability Are the data inside the limits and close to the target?
LSL Target USL

No

Mean Standard deviation Capability Pp Ppk Z.Bench % Out of spec PPM (DPMO)

Comments Before: C1, ... After: C25, ...

Before

Conclusions -- The process standard deviation was not reduced significantly (p > 0.05). -- The process mean changed significantly. It is now further away from the target (p < 0.05). Actual (overall) capability is what the customer experiences.

After

3rd iteration minus 6mls The question was raised at this point if the manufacturer is stating that the machine can operate within +/- 0.7 tolerances why is this not reflected in the process statistics. The reality is that the process has many inputs and the fill tubes and heads alone have not fully resolved the fill height variation issue. This issue needed further investigation but was considered out of scope for this project due to time constraints.

Before/After Capability Comparison for C1, ... vs C31, ... Summary Report
Reduction in % Out of Spec 99% % Out of spec was reduced by 99% from 25.26% to 0.24%. Lower Spec 1970 Customer Requirements Target 2000 Upper Spec 2005

Was the process standard deviation reduced? 0 0.05 0.1 > 0.5

Statistics

Before 2002.5 3.7130 1.57 0.22 0.67 25.26 252650

After 1994.5 3.7130 1.57 0.94 2.82 0.24 2396

Change -8 0 0.00 0.72 2.15 -25.03 -250254

Yes

No
P = 0.500 Did the process mean change?

0.05

0.1

> 0.5

Yes
P = 0.000 Actual (overall) Capability Are the data inside the limits and close to the target?
LSL Target USL

No

Mean Standard deviation Capability Pp Ppk Z.Bench % Out of spec PPM (DPMO)

Comments Before: C1, ... After: C31, ...

Before

Conclusions -- The process standard deviation was not reduced significantly (p > 0.05). -- The process mean changed significantly. It is now further away from the target (p < 0.05). Actual (overall) capability is what the customer experiences.

After

4th iteration minus 8mls It is interesting to note that only 0.24% units will be produced outside the upper specification which is self-imposed while none of the output will fall below the lower specification set by the legal requirement under the packers regulations. This is a far cry from the initial trial that suggested over 65.63% of the output would exceed the upper specification limit.

Before/After Capability Comparison for C1, ... vs C37, ... Summary Report
Reduction in % Out of Spec
100%

% Out of spec was reduced by 100% from 25.26% to 0.04%.

Lower Spec 1970

Customer Requirements Target 2000

Upper Spec 2005

Was the process standard deviation reduced? 0 0.05 0.1 > 0.5

Statistics

Before 2002.5 3.7130 1.57 0.22 0.67 25.26 252650

After 1992.5 3.7130 1.57 1.12 3.36 0.04 391

Change -10 0 0.00 0.90 2.69 -25.23 -252259

Yes

No
P = 0.500 Did the process mean change?

0.05

0.1

> 0.5

Yes
P = 0.000 Actual (overall) Capability Are the data inside the limits and close to the target?
LSL Target USL

No

Mean Standard deviation Capability Pp Ppk Z.Bench % Out of spec PPM (DPMO)

Comments Before: C1, ... After: C37, ...

Before

Conclusions -- The process standard deviation was not reduced significantly (p > 0.05). -- The process mean changed significantly. It is now further away from the target (p < 0.05). Actual (overall) capability is what the customer experiences.

After

5th iteration minus 10mls I think its very unlikely that the organisation will sanction a move of this magnitude. It is interesting to note that this is the first simulation to report a Ppk value greater than 1.0. It also starts to look like the target value of 2000mls is the upper specification limit. Implementation and Verification While the performance of the filler is crucial it is not the only process driver we identified in the analysis phase. A key component of our improvement strategy was operator training which included a range of issues that could potentially impact on raw material usage. It is difficult to isolate any one aspect of these activities but a key focus has been placed on the operating settings of the filler and the on-line monitoring of the output by the operator. Assessing Benefits of the Proposed Solution

Raw Material Yield based on output


Week No. WK 12 WK 13 WK 14 Line 2 % yield 99.8 100.2 99.9 Month March 2013 April 2013

The latest and most up to date raw material yield figures post improvement are very encouraging and would suggest a yield improvement from 99.1 to 99.96 a yield percentage improvement of 0.86. As we indicated earlier it is estimated that each .1 improvement generates an annual saving of 15,000. On this basis is reasonable to assume that this six sigma project could generate savings in the region of 129,000 on an annualised basis. Experience & Conclusions: While considerable progress has been made and the time and money invested by the organisation in this project will make a return. I cant help feeling a little disappointed for the team. My disappointment is focused on the post improvement performance of the filler. If you compare the manufacturers specifications of +/- .7% to the process standard deviation +/- 3 (3.713 x 3 = +/11.139) for fill heights. I have no doubt that the engineering team will address the filler performance issue going forward and the operators will continue to optimise the process output. On the positive side the improvements has been well received by the project sponsor and the wider organisation.

You might also like